Note To Users

NOTE TO USERS
Page(s) not included in the original manuscript and are

unavailable from the author or university. The manuscript
was scanned as received.
193-242
This reproduction is the best copy available.
UMI
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
EVALUATING THE EFFECT OF MERIT AID AS A HIGHER EDUCATION
POLICY TOOL USING TIME SERIES ANALYSIS
by
Michelle Johanna Nilson
Bachelor of Arts
Wayne State University
1998
Master of Arts
New Mexico State University
2003
A dissertation submitted in partial fulfillment
of the requirements for the
Doctor of Philosophy Degree in Higher Education Administration
Department of Educational Leadership
College of Education
Graduate College
University of Nevada, Las Vegas
August 2006
UMI Number: 3244002
Copyright 2006 by
Nilson, Michelle Johanna
All rights reserved.
INFORMATION TO USERS
The quality of this reproduction is dependent upon the quality of the copy
submitted. Broken or indistinct print, colored or poor quality illustrations and
photographs, print bleed-through, substandard margins, and improper
alignment can adversely affect reproduction.
In the unlikely event that the author did not send a complete manuscript
and there are missing pages, these will be noted. Also, if unauthorized
copyright material had to be removed, a note will indicate the deletion.
UMI
UMI Microform 3244002
Copyright 2007 by ProQuest Information and Learning Company.
All rights reserved. This microform edition is protected against
unauthorized copying under Title 17, United States Code.
ProQuest Information and Learning Company
300 North Zeeb Road
P.O. Box 1346
Ann Arbor, Ml 48106-1346
Copyright by Michelle J. Nilson 2006
All Rights Reserved
Dissertation Approval
The Graduate College
July 24 20 06
The Dissertation prepared by
M ichel le J . N i l so n
Entitled
_______Evaluating the E f f e c t o f Merit Aid as a Higher Education P o l i c y
______________________ Tool Using Time S e r i e s An alysis_______________________
is approved in partial fulfillment of the requirements for the degree of
___________ Doctor of Philosophy i n Educational Leadership
Examination Committee Chai\
Dean of the Graduate College
Committee Member
Examinat iam/Conun it t/e Member
Graduate College Faculty Representative
1017-52 i f
UNLV
ABSTRACT
Evaluating the Effect of Merit Aid as a Higher Education
Policy Tool Using Time Series Analysis
by
Dr. Mario Martinez, Examination Committee Chair
Associate Professor of Higher Education
Since 1991, seventeen states have dramatically altered the criteria they use to
distribute student financial aid to include a larger proportion of merit-based awards.
Across these states, the amounts and lengths of the awards vary. There are states, such as
Georgia and Florida, which provide full tuition and fees for four or five years, depending
on the program of study. In sharp contrast, Michigans Merit Award Scholarship is a one
time $2,500 award. States also differ in terms of the selection criteria used to award
merit aid.
Using a quasi-experimental interrupted and pooled time series design derived
from research on public budgeting, this research investigates the linkage between merit
aid and participation by sector (public and private) and level (two- and four-year) in
states and institutions. In this study, merit aid is characterized into three main categories:
full tuition, partial tuition, and one-time payment of awards. Interrupted time series is
applied to these three categories, to discern whether there are differences in merit aid
programs and their effects on enrollment by sector and level. In addition, pooled time
iii
series analysis is utilized to examine the effects of these programs by aid category across
states, sectors and levels.
Findings indicate that the adoption of a merit aid policy significantly changed
enrollment in 9 of the 15 states investigated in this study. While the results were mixed,
generally, states experienced a greater long term positive effect than negative or short
term effects, indicating that more students take advantage of the programs over time.
Full tuition payment policies had a short term significant effect on enrollment in the 4-
year public sector analysis. However, partial tuition payment policies had a significant
positive long and short term effect on the 2-year public sector. This finding supports
earlier work that theorized that merit aid encourages students who might not otherwise
enroll to do so.
IV
TABLE OF CONTENTS
ABSTRACT..................................................................................................................................iii
LIST OF FIGURES....................................................................................................................vii
CHAPTER 1 INTRODUCTION............................................................................................ 1
Statement of the Problem....................................................................................................... 3
Purpose of the Research......................................................................................................... 3
Significance of the Study....................................................................................................... 4
Research Questions.................................................................................................................6
Limitations of the Study......................................................................................................... 7
Definition of Terms............................................................................................................... 10
CHAPTER 2 LITERATURE REVIEW................................................................................ 12
Student Financial Aid...........................................................................................................13
Brief History of Financial Aid in the United States..........................................................14
Federal Student Financial Aid ............................................................................................ 16
Institutional Student Financial Aid.....................................................................................17
Institutional Responses to State and Federal Aid .............................................................19
State Merit Based Student Financial Aid........................................................................... 19
Postsecondary Participation................................................................................................ 24
CHAPTER 3 METHODOLOGY........................................................................................... 28
Introduction............................................................................................................................28
Main Methods....................................................................................................................... 28
Sources of Error.................................................................................................................... 30
Pooled Time Series Analysis.............................................................................................. 30
Levels of Analysis.................................................................................................................31
Hypotheses.............................................................................................................................33
Description of Analyses....................................................................................................... 39
Impact Analysis.................................................................................................................... 41
Modeling Using Regression................................................................................................ 42
Data......................................................................................................................................... 44
Summary................................................................................................................................ 46
CHAPTER 4 STATE LEVEL ANALYSIS RESULTS......................................................47
Introduction............................................................................................................................47
Study Results........................................................................................................................ 49
Research Question 1 ............................................................................................................ 49
Research Question 2 .............................................................................................................82
v
Summary ..................................................................................................................................... 87
CHAPTER 5 SECTOR LEVEL ANALYSIS RESULTS...................................................89
Introduction............................................................................................................................ 89
Research Question 3 ............................................................................................................ 90
Research Question 4 ...........................................................................................................151
Summary ............................................................................................................................ 170
CHAPTER 6 SUMMARY, CONCLUSIONS, AND RECOMMENDATIONS............173
Summary of Results............................................................................................................173
Comparison with Existing Literature............................................................................... 187
Implications for Policy....................................................................................................... 193
Implications for Future Research...................................................................................... 190
APPENDIX................................................................................................................. on CD Rom
State Merit Aid Program Summaries............................................................................... 193
Methodology Matrix...........................................................................................................197
State Summary Research Question 1............................................................................... 205
State Summary Research Question 3 ............................................................................... 207
Omitted Cases..................................................................................................................... 214
REFERENCES..........................................................................................................................243
VITA........................................................................................................................................... 252
vi
LIST OF FIGURES
Figure 1 Estimated student aid by source...............................................................................19
Figure 2 Impact Analysis........................................................................................................ 42
Figure 3 Annual Enrollment Pre- and Post Merit Aid Adoption, Alaska........................ 50
Figure 4 Annual Enrollment Pre- and Post- Merit Aid Adoption, Arkansas................. 52
Figure 5 Annual Enrollment Pre- and Post- Merit Aid Adoption, Florida......................55
Figure 6 Annual Enrollment Pre- and Post- Merit Aid Adoption, Georgia....................57
Figure 7 Annual Enrollment Pre- and Post- Merit Aid Adoption, Kentucky................. 59
Figure 8 Annual Enrollment Pre- and Post- Merit Aid Adoption, Louisiana................ 61
Figure 9 Annual Enrollment Pre- and Post- Merit Aid Adoption, Michigan................. 64
Figure 10 Annual Enrollment Pre- and Post- Merit Aid Adoption, Missouri.................. 66
Figure 11 Annual Enrollment Pre- and Post- Merit Aid Adoption, Mississippi.............. 68
Figure 12 Annual Enrollment Pre- and Post- Merit Aid Adoption, New Mexico........... 70
Figure 13 Annual Enrollment Pre- and Post- Merit Aid Adoption, Nevada.....................73
Figure 14 Annual Enrollment Pre- and Post- Merit Aid Adoption, South Carolina........75
Figure 15 Annual Enrollment Pre- and Post- Merit Aid Adoption, South Carolina........77
Figure 15 Annual Enrollment Pre- and Post- Merit Aid Adoption, Tennessee............... 79
Figure 16 Annual Enrollment Pre- and Post- Merit Aid Adoption, Washington............. 80
Figure 17 Annual Enrollment Pre- and Post- Merit Aid Adoption, West Virginia..........81
vii
LIST OF TABLES
Table 1 Regression Analysis Coefficients, Final Model for Alaska...............................51
Table 2 Regression Analysis Coefficients, Final Model for Arkansas.......................... 53
Table 3 Regression Analysis Coefficients, Final Model for Florida..............................56
Table 4 Regression Analysis Coefficients, Final Model for Georgia.............................58
Table 5 Regression Analysis Coefficients, Final Model for Kentucky......................... 60
Table 6 Regression Analysis Coefficients, Final Model for Louisiana.......................... 62
Table 7 Regression Analysis Coefficients, Final Model for Michigan............................65
Table 8 Regression Analysis Coefficients, Final Model for Missouri.............................67
Table 9 Regression Analysis Coefficients, Final Model for Mississippi........................ 69
Table 10 Regression Analysis Coefficients, Final Model for New Mexico..................... 71
Table 11 Regression Analysis Coefficients, Final Model for Nevada............................... 74
Table 12 Regression Analysis Coefficients Final Model for South Carolina..................76
Table 13 Regression Analysis Coefficients Final Model for South Carolina..................78
Table 14 Regression Analysis Coefficients Final Model for Full Tuition....................... 83
Table 15 Regression Analysis Coefficients Final Model for Partial Tuition................... 85
Table 16 Regression Analysis Coefficients Final Model for One time payment........... 87
Table 17 Regression Analysis Coefficients Final Model for Alaska,
2-Year Pubic.............................................................................................................91
Table 18 Regression Analysis Coefficients Final Model for Alaska,
4-Year Private.......................................................................................................... 93
Table 19 Regression Analysis Coefficients Final Model for Arkansas,
2-Year Public........................................................................................................... 94
4-Year Public........................................................................................................... 95
2-Year Private.......................................................................................................... 97
4-Year Private.......................................................................................................... 98
Table 23 Regression Analysis Coefficients Final Model for Florida,
2-Year Public........................................................................................................... 99
4-Year Public..........................................................................................................100
4-Year Private........................................................................................................ 102
Table 26 Regression Analysis Coefficients Final Model for Georgia,
4-Year Public..........................................................................................................103
4-Year Public..........................................................................................................104
viii
4-Year Private........................................................................................................ 106
Table 29 Regression Analysis Coefficients Final Model for, Kentucky
2-Year Public..........................................................................................................107
Table 30 Regression Analysis Coefficients Final Model for Kentucky,
4-Year Public..........................................................................................................109
Table 31 Regression Analysis Coefficients Final Model for Kentucky,
4-Year Private.........................................................................................................110
Table 32 Regression Analysis Coefficients Final Model for Louisiana,
2-Year Public..........................................................................................................I l l
4-Year Public..........................................................................................................113
2-Year Private........................................................................................................ 114
4-Year Private........................................................................................................ 115
Table 36 Regression Analysis Coefficients Final Model for Michigan,
2-Year Public..........................................................................................................116
4-Year Public..........................................................................................................118
2-Y ear Private........................................................................................................ 119
4-Year Private........................................................................................................ 121
Table 40 Regression Analysis Coefficients Final Model for Missouri,
2-Year Public......................................................................................................... 122
Table 41 Regression Analysis Coefficients Final Model for Missouri,
4-Year Public......................................................................................................... 123
Table 42 Regression Analysis Coefficients Final Model for Mississippi,
2-Year Public......................................................................................................... 125
4-Y ear Public..........................................................................................................126
2-Year Private........................................................................................................ 127
4-Year Private........................................................................................................ 129
Table 46 Regression Analysis Coefficients Final Model for New Mexico,
2-Y ear Public......................................................................................................... 130
4-Y ear Public..........................................................................................................131
2-Y ear Private........................................................................................................ 132
Table 49 Regression Analysis Coefficients Final Model for Nevada,
4-Year Public......................................................................................................... 134
Table 50 Regression Analysis Coefficients Final Model for Nevada,
4-Y ear Private........................................................................................................ 135
ix
Table 51 Regression Analysis Coefficients Final Model for South Carolina,
2-Year Public......................................................................................................... 136
2-Y ear Private........................................................................................................ 138
4-Y ear Private........................................................................................................ 139
Table 54 Regression Analysis Coefficients Final Model for Tennessee
2-Year Public......................................................................................................... 140
Table 55 Regression Analysis Coefficients Final Model for Tennessee,
2-Y ear Private........................................................................................................ 142
Table 56 Regression Analysis Coefficients Final Model for Tennessee,
4-Y ear Private........................................................................................................ 143
Table 57 Regression Analysis Coefficients Final Model for Washington,
2-Year Public......................................................................................................... 145
4-Y ear Public......................................................................................................... 146
4-Year Private........................................................................................................ 148
Table 60 Regression Analysis Coefficients Final Model for West Virginia,
2-Y ear Public......................................................................................................... 149
Table 61 Regression Analysis Coefficients Final Model for West Virginia,
4-Y ear Public......................................................................................................... 150
Table 62 Regression Analysis Coefficients Final Model for Full tuition,
2-Y ear Public......................................................................................................... 152
4-Year Public..........................................................................................................154
2-Y ear Private........................................................................................................ 155
4-Y ear Private........................................................................................................ 157
Table 66 Regression Analysis Coefficients Final Model for Partial tuition,
2-Year Public......................................................................................................... 159
4-Y ear Public..........................................................................................................160
2-Y ear Private........................................................................................................ 162
4-Y ear Private........................................................................................................ 163
Table 70 Regression Analysis Coefficients Final Model for One time payment,
2-Year Public......................................................................................................... 165
4-Year Pubic...........................................................................................................166
2-Year Private........................................................................................................ 168
x
4-Y ear Private........................................................................................................ 169
Table 74 Regression Analysis Coefficients Impact of Policy Adoption by Sector....... 184
Table 75 Regression Analysis Coefficients Impact of Policy Type by Sector...............187
xi
ACKNOWLEDGMENTS
This dissertation is dedicated to the men and women of the Henderson Police
Department in Henderson, Nevada. I would especially like to thank Officer David
Adams.
Thank you to Mario Martinez for being such a brilliant mentor, advisor, and
dissertation chair. Thank you for your continued guidance on all aspects of life,
scholarship, and work. Thank you, Mimi Wolverton (SWMBO), for pushing me harder
than I ever thought I could be pushed and for opening more doors than I ever imagined
existed. Thank you to Bob Ackerman (my Chief) for being a sounding board for so much
of this experience and for being so generous with your wisdom and humor. Thank you to
Chris Stream for your wonderfully contagious enthusiasm and optimism (and for
introducing me to the healthy bowl). Also, I would like to thank Chad Cross for taking
the time to review my statistical analysis and Robert McCord for being a great neighbor.
This dissertation was heavily influenced by my internship at ECS with Sandy
Ruppert, who taught me the importance of having a life outside of work and the value of
being able to merge practice with theory. Thank you to Carl Freeman, Amy Weibel,
Maria Ferriera, Regina Zibuck for encouraging me to pursue a Ph.D. in the first place.
There are a number of people and organizations that helped and supported me
along the way that I would like to thank: the faculty at NMSU, especially Dana
Christman and her student, Susan Cardenas. Thank you to Colleen OBrien and the Pell
Institute for their generous support of my attendance at the Student Financial Aid
xii
Research Network Conference in 2005. Thank you to the Graduate and Professional
Student Association at UNLV for supporting my research and travel. Also, thank you to
Jennifer Blatz at the Harvard Civil Rights Project for providing copies of the two CRP
publications on merit aid. And a hearty thank you to Phil Beuth for all of his hard work.
My classmates, colleagues, friends and the faculty and staff at UNLV, in Las
Vegas, and back in Michigan have been hugely important in my growth and development
as a person and a scholar. They reinforced, challenged, and taught me a deeper
understanding of the value of thoughtfulness, kindness, courage, compassion, integrity,
family, diversity, and reminded me of the beauty and power of friendship- and for that I
am eternally grateful to them.
And finally, I would like to thank my family, who encouraged me to pursue my
passions, where ever they led. Especially thank you to my Mom & Sepp and my Dad,
words are not enough to express my gratitude, my actions will have to speak loudly.
Much love and thanks to my sisters Suzy, Nicole, Tanya, Whitney and my brother, Seppi.
Michael, Logan, Sadie, Loma, Riley, Evamarie and Aspen- thank you for sharing your
bright new worlds with me. And finally, to my surrogate families, the Bakers, the
Gallants, and the Sorensens thank you for taking me in and sharing so many wonderful
memories.
xiii
CHAPTER 1
INTRODUCTION
The principle of rewarding students for their success and hard work is not a novel
idea. High performing students are rewarded with opportunities to participate in many
activities from attendance at exclusive and highly selective schools, honors programs,
and internships, to receiving private scholarships to support their endeavors. A recent
state policy development in the reward system for high achieving students is merit based
financial aid for postsecondary students. The need for continual, additional and new
approaches that examine merit aid and its effects on participation is heightened by the
tendency of states to copy policies from other states. Legislators, in particular in states
surrounding Georgia, adopted similar merit aid policies as Georgias 1993 first broadly
available state merit aid program, in hopes of reaping some of the same benefits that they
saw in Georgia (Cohen-Vogel & Ingle, 2006; McLendon, Heller, & Young 2005).
A recent examination by the Education Commission of the States found that since
1993, seventeen states have begun to reward meritorious students with financial aid
(ECS, 2005). However, prior to 1993, the main criteria for student financial aid awards
were financial need, with less than 10 percent of grant dollars going to merit aid (Heller,
2004). The three main purposes behind this legislation are to: 1) keep the best students in
the state, 2) reward and encourage their hard work, and 3) promote college access and
attainment (Heller, 2002).
1
Although states may emulate policies from each other, merit aid is a term that
now encompasses a wide variety of student financial aid policies enacted across the
states. On one end of the spectrum, there are states that award full tuition, fees, and even
book money, as in the case of the Georgia HOPE scholarship. On the other end of the
spectrum is Michigans Merit Scholarship award, which is a one-time $2500 payment.
Between these two extremes are several programs, such as the ones in Nevada, Missouri,
and Mississippi, which pay on average up to $2,500 per year, with lifetime award limits
of around $ 10,000 each.
There are relatively few studies that examine the full impact of these different
types of policies on participation across all of the states, in part due to their relatively new
arrival on the public policy scene. In a recent study using Census 2000 data from seven
Southern states merit aid programs, Dynarski (2003) found that the introduction of a
merit aid program increased the probably of enrolling in postsecondary education by 5 to
7 percentage points. Each of the states in her study was in the South and they all had
similar policies regarding the amounts of aid.
Other studies, which examine individual state programs, have met with mixed
results. For example, Cornwell and Mustard (2003b) found that in states where tuition,
student fees, and books are paid for, there was a significant effect on enrollment in the
state; Heller (2003) found that in Michigan, where the award is limited, the incentive
effects are marginal. However, the wide variety of merit aid programs and their different
impacts on participation by sector (public and private) and level (two- and four-year)
remains unexamined. In her 2004 dissertation, Patricia Farrell conducted a
2
comprehensive evaluation of existing merit aid policies at the state level and
recommended further study of their impact on institutions, levels, and sectors.
Statement of the Problem
Given that merit aid has not provided a clear incentive effect in all of the states
where it was adopted, there is some debate as to whether policy makers should look to
merit financial aid programs to provide incentives for students. One issue that remains is
that, upon examination of the debates about merit aid, there is little differentiation
between types of merit aid programs and their corresponding effects on enrollment over
time. This study is a longitudinal investigation of the impact of the different types of
merit aid programs and their impact on enrollment by sector and level.
The literature on public policy can be instrumental in building an understanding
of long term state finance policy changes. Public policy scholars routinely examine large
scale interventions, such as programs similar to the merit aid grants. One such technique
that has been useful in determining the impact of an intervention is the use of time series
analysis. This type of post-hoc analysis is a longitudinal examination of patterns before
and after an intervention, often using secondary data sets. Time series analysis has a long
history of use in economics, public health and epidemiology, agriculture sciences,
psychology, business, and public policy (Taggart, 1989).
Purpose of the Research
The purpose of this study is to examine the impact of merit aid on first time
freshman enrollment by sector (private, non-profit and public institutions) and level (2-
3
year and 4-year institutions) in the twelve states that have had the policy for more than
five years. Since policymakers in neighboring states often adopt policies that are similar,
the merit aid policies were characterized into three types:
1. Full tuition and benefits;
2. Partial tuition for entire program; and
3. One time payment of partial tuition
The impact of these policies on institutional and state enrollment patterns was
examined. Characterizing the types of merit aid programs serves to isolate and clarify the
impact each of the three main types of programs commonly found in the states. In
addition, the three types of merit aid policy were compared between types of programs to
determine the relative impact on participation. For example, the full tuition and benefits
programs were compared to the partial tuition and one time payment for partial tuition
programs to determine if one of the programs had a larger impact on enrollment.
Significance of the Study
In a recent editorial article, Smart (2005) articulated the attributes of exemplary
quantitative research. One of the suggestions Smart has for young scholars is that they
borrow the best examples of theoretical and methodological paradigms of other
disciplines to important topics on the higher education research agenda (p 465). This
methodological technique is currently utilized largely in public policy (Holmes, M. D.,
Daudistel, H. C. & Taggart, W. A., 1992; Taggart, 1989) and economics; here it is
transferred to higher education policy in order to analyze merit aid policies across the
states. Time series analysis is not a new technique; it is a methodology that dates back to
4
early agricultural studies and has roots that date back over 400 years (Klein, 1997). This
approach is novel in that there are few studies that utilize quasi-experimental time series
techniques in higher education policy analysis.
One final suggestion that Smart has is that research have important implications
for future research and practice and policy. From this time series analysisand
depending on the results of the analysisa model that can be used in states to project
their enrollments based on the type of merit aid policy that is adopted will be constructed.
In addition, this study will inform future research using time series analysis as well as
higher education policy by contributing to the research using a proven cross-disciplinary
technique.
The study contributes to the knowledge of merit aid and its effect on enrollment in
yet a third way. Many authors have investigated from a broad, state perspective, the
effect of tuition increases or grant amounts on enrollment. The work of Dynarski (2001)
looked at the impact of grants on enrollment. Others have proposed standardized price
response coefficients (SPRCs), which provides an elasticity measure, a change in the
probability of enrollment for every $100 change in net price (e.g., Jackson &
Weathersby, 1975; Leslie & Brinkman, 1988; McPherson, 1978) (St. John, Asker, &
Hu, 2001 in Paulsen & Smart, 2001.) Tierney (1980), found that a $100 increase in
grant offers by private relative to public institutions would increase the probability that
such a student matriculates at a private institution to .67 (p. 541). By examining the
impact of merit aid on enrollment, this research was able to determine whether significant
differences in participation exist across programs, which may provide insight into the
elasticity of merit aid on enrollment.
5
Research Questions
The gaps in the research raise several questions about merit aid programs, and
their associated policy implications. The following research questions guided the data
collection and analysis for the study:
State Level Analysis
1) How has merit aid impacted non-profit postsecondary participation in each of the
states that has adopted this student financial aid policy?
2) What is the effect of the following three types of merit aid programs on state-level
participation:
a. Full-tuition programs
b. Partial tuition programs
c. One time payment programs
Sector (Public and Private) and Level (Two- and Four-Year!
3) How has merit aid impacted participation by sector and level in the states that
have adopted this financial aid policy?
4) What is the effect on enrollment by type of aid program in each sector and level
within each state?
a. Each state
i. 2-year public
ii. 2-year private
iii. 4-year public
iv. 4-year private
6
Limitations of the Study
The following outlines some of the limitations of this study.
Aggregate and self reported datathe aggregate participation data that were
used in this study were collected from the National Center for Education
Statistics IPEDS data set. As such, the data are self-reported by the
institutions in each state and can be incomplete or omit institutions that are in
the states selected for this study.
Self selectionas in almost all studies that involve higher education,
participation was not compulsory and therefore reflects a self-selecting sample
of the larger population. As such, care must be taken when attempting to
generalize to the whole population.
Autocorrelationthere is a high correlation between current freshmen
enrollment and enrollment in the higher levels. In most states there has been
an increase in the number of students that are enrolling in postsecondary
education over the past 20 years (Mortenson, 2006). This increasing trend
toward enrollment can lead to autocorrelation in enrollment from one year to
the next. The first way this study limits the effects of this autocorrelation is to
limit the study to the impact of first time undergraduates only. A test that is
used to determine i f there is autocorrelation is to check the Durbin-Watson
statistic for each model that was developed from the time series regression
equations, which is fully explained in Chapter 3 (Meier and Brudney, 2002).
Alternative explanationsan assumption of this study was that the change in
enrollment at the time of the policy implementation was due to the policy
7
effects rather than some other explanation. Alternative explanations for
increased enrollment can be ruled out by increasing the number of cases,
which is why this study examines similar programs across several states. In
addition, by looking at the effects of the policy across institutional sectors and
levels in several states, the number of cases is increased to a point where
outliers can be identified and explained.
This study does not differentiate between those students who are residents of
merit aid states versus those who are not; it was a broad examination of the
impact of the policy adoption on all first time undergraduate enrollments
across the state.
Multicollinearity- due to the nature of time series analysis and the use of
dummy variables, there were instances where multicollinearity was an issue.
For example, both the long term impact (TTT) and time (T) trend variables
were used to determine trends and used counting variables that overlapped in
their measures. Another example is the high level of correlation between the
other independent variables, such as the number of high school graduates
(HSGRAD) and higher education appropriations (HEAPP).
Once a person determines that he or she wants to participate in higher education,
they face the task of selecting an institution and getting accepted. At each stage of the
process between determining that one wants to go to college and getting in the chair on
the first day of class, there are factors that weigh into the decisions along that path. The
college choice, participation, enrollment, matriculation, and participation literature has
identified several of those factors: parent attainment, socioeconomic status, gender,
8
race/ethnicity, financial aid, self efficacy, grade point average, standardized test scores,
and tuition, (Cabrera & LaNasa, 2000; Leslie & Brinkman, 1988; St. John, 1990) to
name just a few. While this study is about financial aid, which is one factor of many in
the decision to enroll, this study does not take into account those individual differences
that can play a role in a persons decision to enroll. As such, caution should be used
when trying to generalize this study and its results to the decision of a single person.
Since the merit aid in states can only to be used in not for-profit postsecondary
institutions, this study examines only the non-profit sector of higher education in the
United States.
Van der Klaauw (2002) theorizes that college administrators have little
knowledge about the alternative enticements that are available to students. These
alternatives can include campus job opportunities, special programming, financial
aid offers from other institutions, military, external job opportunities, and a host
of other omitted variables that may play a role in enrollment decisions. He argues
that since financial aid is a part of the enrollment decision process, it is an
endogenous, or nested, variable. In this study, each of the states postsecondary
financial aid is awarded to students based on a given set of criterion that students
are generally aware of ahead of time and they can determine for themselves
whether or not they qualify for merit for aid. While this study does not control for
institutional or federal financial aid, the impact of state aid on enrollment is
reasonably isolated by examining only similar state financial aid programs.
9
Definition of Terms
For the purposes of this study, the following definitions are for clarity and to
ensure a common understanding of terms:
College choice: The decision process and conclusion concerning where one
should matriculate.
College selection: Part of the decision process where a student candidate chooses
to enroll from among a variety of institutions.
Enrollment: The measure for postsecondary participation, also referred to as
demand as represented by those choosing to enroll in postsecondary education (Buss,
Parker, & Rivenburg, 2004).
Integrated Postsecondary Education Data System (IPEDS): A national database
on postsecondary education that is housed at the National Center for Education Statistics.
They are institutionally reported data that contain enrollment information.
Level: Describes whether a postsecondary institution offers 2-year, 4-year or
technical degree programs.
Matriculation: The act of being enrolled as a student in a postsecondary
educational institution.
Merit-based aid: In this case, it refers to state supported student financial aid
whose primary criterion for award is merit rather than need.
Participation gap: The difference between the enrollment at the current rate
(baseline rate) and enrollment at some higher rate set by the benchmark states (Martinez,
2004).
10
Participation rate: the number of students in postsecondary education divided by
the total population (Martinez, 2004).
Postsecondary participation: Refers to a person attending a public or private
degree-granting college or university (Census, 2000).
Sector: Refers to either public or private nonprofit postsecondary institutions
Factors of merit aid programs: The three main factors of merit aid programs are
award criteria for selection, length of award, and amount of award.
Criteria for selection: Refers to the criteria used to determine award eligibility,
typically grade point average, standardized test score, or class rank.
Length of award: Refers to the length of time of the award (semesters).
Amount of award: Refers to the award dollar amount per year.
Categories of merit aid programs: Refers to the broad categorization of merit aid
programs by award amount (either full tuition, partial, or one-time payment).
Full tuition payment programs: Refers to merit aid student financial aid programs
that pay for all postsecondary tuition expenses.
Partial tuition payment programs: Refers to merit-based student financial aid
programs that pay a recurring award for some portion less than 100% of postsecondary
tuition expenses.
One-time payment programs: Merit aid programs that make a one time payment
to students for their performance.
11
CHAPTER 2
LITERATURE REVIEW
This chapter is a review the literature on student financial aid and postsecondary
participation. State merit aid is the focus of this study and the literature review. The
initial introduction to the broad issues that underlie financial aid is followed by a brief
history of student financial aid in the United States. While this study focuses on state
merit aid programs and their impact on enrollment, it is also important to understand both
federal and institutional financial aid structures. In addition to providing a fuller
understanding of the existing structures, an understanding of federal and institutional
financial aid programs provides context for the state merit aid program structures that are
currently in place. The balance of the chapter reviews the current literature on state merit
aid programs. Finally, a brief review of the public policy perspectives related to large
scale state-wide intervention programs, such as the merit aid programs is provided.
Throughout the reviewed literature cited in this chapter, the underlying theme is
the interaction between financial aid and its impact on participation. Where it is
appropriate, the interaction between the aid and participation variables will be reviewed
in the financial aid literature; where possible, the literature review on participation will be
covered separately. Additional issues related to aid and participation, such as college
choice, will be discussed as appropriate.
12
Student Financial Aid
There are several factors that influence the decision about whether or not a person
will go to college. Several researchers (Hossler, Braxton, & Coopersmith, 1989; Hossler
& Gallagher, 1987; Hossler, Schmit & Vesper, 1999; Kim, 2004; McDonough, 1997;
Tierney, 1980) have found that financial aid plays a significant role in the choice of
colleges for students. Cabrera and La Nasa (2000) proposed a model for the college
choice processes that showed that several factors influence college choice. Among those
factors were parental characteristics (such as education, income, and occupation,
collegiate experiences, encouragement, and involvement), student characteristics (ability,
qualifications, aspirations), and institutional or sector factors (such as availability of
information, features of potential institutions, and cost of attendance and financial aid)
(Cabrera & La Nasa, 2000). Of all of the factors identified as playing a role in student
participation, institutions, governments, and administrators generally have broad
influence over only two general areas of student participation in higher education
admissions and financial aid (Van der Klaauw, 2002). Pema and Titus (2004) add to the
list of tools within the scope of higher education policy controltuition, appropriations
to institutions, and policies related to K-12 preparation for higher education.
Researchers typically examine state and federal financial aid as separate entities
the results of which have been mixed. The research on the impact of institutional
financial aid has found that it has a significant effect on participation (Jackson, 1978;
Lumina, 2003, 2004). Research on state aid has found that it has a significant impact on
sectors of the college going population but the size of that impact is mixed, depending
largely on the amount of award (Heller, 1999; Dynarski, 2000; Kane, 2003). The federal
13
financial aid research related to student participation has shown a negligible effect at best
(Maag & Fitzpatrick, 2004; Dynarski, 2003; Heller, 1999). Further, most studies find
that state financial aid plays a significant role in the decision making process of students
(Heller, 1997; St. John, et al., 2004). The impact of state aid is particularly significant
when examining students with low socioeconomic status (Alexander, 2001) or those of
African American and Hispanic origins (Heller, 1997; Kim, 2004; Stewart & Post, 1990).
In addition, Seftor and Turner (2002) found that the federal Pell grant played a significant
effect on the enrollment of adults in higher education. There remain gaps in the literature
concerning types of financial aid programs and their impact on participation. In addition,
because state-wide merit based financial aid programs are relatively new policy
instruments, there is little research available on their effectiveness or impact on student
participation.
Brief History of Financial Aid in the United States
While there was institutional student aid grant in place a short seven years after
the establishment of Harvard University, the first federal student aid program in the US
was not established until nearly three hundred years later, in 1935. Under the WPA, the
National Youth Administration (NYA) provided part time jobs for over 700,000 college
students. This initiative was started as part of the New Deal with the intention of
providing jobs to youth during a time of economic depression. In its later years, it
changed its focus to encompass job training for youth assisting in the war effort and was
eventually abolished in 1943 (The Eleanor Roosevelt Papers).
14
Shortly after the abolition of the NYA, federal aid was extended to servicemen
through the Servicemans Readjustment Act of 1944, or GI Bill of Rights. This was the
first large scale legislation that allocated money to individuals rather than institutions.
Gladieux (2003) called the GI Bill the most important education legislation of the 20th
century in the US. In addition, he credits the GI Bill with inspiring efforts to broaden
higher education access through subsequent legislation such as the Higher Education Act
of 1965 and the Pell Grant, passed in the reauthorization of 1972.
Access continued as a goal throughout the 1970s on into the new millennium;
however, student-financing policy shifted from grants to loans over the past 30 years (St.
John, 1994). In 1974, the federal government established the Student Loan Marketing
Association (Sallie Mae) and in 1976, the Reauthorization of the Higher Education Act
provided for state loan-guarantee agenciesboth of which would play a major role in the
changing landscape of higher education finance for the next 30 years. The legislation of
the 1990s provided enhanced access to loans for students and parents paying for higher
education. The 1997 Taxpayer Relief Act provided for deductions to taxpayers for higher
education interest on loans, provided for Lifelong Learning tax credits, Education IRAs,
section 529 plans, Federal Hope Scholarship, and tax exclusions for employer paid
contributions to education (FinAid, 2005).
One of the current issues in federal financial aid circles surrounds lending
policies. It is of concern for many reasons. According to Gladieux and Pema (2005),
lending has increased to nearly 50 percent of all entering freshmen and to nearly two-
thirds of those entering four-year colleges. Perhaps most disturbing, nearly one-fifth of
15
those students drop out with significant loan debt and nearly one-fourth of borrowers who
dropped out defaulted on at least one loan. (Gladieux & Pema, 2005).
Federal Student Financial Aid
There are three main ways in which the federal government provides financial aid
to students: loans, grants, and work-study. According to Martinez (2005), studies on the
impact of federal financial aid on participation focus primarily on two main questions: 1)
Who participates? And 2) Where do they participate in higher education? These
questions are intermingled in the research through various studies of choice, sector
participation, and participation by race and gender. St. John (1991) conducted a review
of the research literature and found that student aid programs are an effective mechanism
for encouraging equal educational opportunity but that reductions in federal grant dollars
have contributed to the participation gap of minority students to higher education. A
more recent study by Ruppert (2003) supports the finding that there is a persistent gap in
minority participation in higher education.
The impact of student financial aid on enrollment is mixed, largely because of the
different types of aid that are available. Jacksons 1978 study of grant aid and student
enrollment found that students awarded $100 more aid were 0.1 percentage points more
likely to attend than otherwise similar nonapplicants. He goes on to state that his data
imply that, the award of aid is more important than the amount, although he does warn
against the generalizability of this finding to other students who received dissimilar
amounts of awards. Heller (1997,1999) and Leslie and Brinkman (1988) confirmed that
students respond to financial grant aid awards and tuition pricing and that the responses
16
vary by race and ethnicity. A more recent study by Dynarski (2001) on the elimination of
the Social Security Student Benefits Program found that offering $1,000 ($1997) of
grant aid increases educational attainment by .20 years and the probability of attending
college by five percentage points.
Using the National Postsecondary Education Student Aid Survey of 1986-87, St.
John & Starkey (1995) found that certain types of grant aid were negatively associated
with persistence for low-income students. In addition, they found that, the amount of
work study awarded was significantly and negatively associated with persistence by
lower-middle-income students. St. John and Starkey argue that the cause may be that,
the average grant award was apparently insufficient relative to the average tuition charge
facing the low-income student (p. 173). Similarly, Spaulding (2003) found that at the
University of Washington, federal student loans and federal work-study had a negative
influence on the students decision to enroll. This combination of federal student loans
and work-study is referred to as student self help. The reliance on these policies of
student self help has disparate impact on low SES students, who are eligible for work-
study programs. The benefits of work-study, as demonstrated by increased involvement
on campus, retention and the like must be supported by additional grant funding rather
than loans in order to encourage these students to persist through to graduation.
Institutional Student Financial Aid
The first recorded financial aid is a scholarship established by Lady Anne
Radcliffe Mowlson in 1643 at Harvard University (FinAid, 2005). Since that time,
institutional aid has grown to nearly $23 Billion, or nearly one quarter of all financial aid
17
granted to students (see Figure 1 below). One of the tools that institutions have to
influence the composition of an incoming freshman class is the financial aid packages it
has to offers students (Thistlethwaite, 1958; Epple, Romano, & Sieg, 2002; van der
Klaauw, 2002). Typically institutional financial aid is used to get a subset of those
admitted, especially those with the greatest academic ability, to enroll.
Private grants, scholarships, and loans are a category that is truly impossible to
determine the full amount of private financial aid distributed nationally to students.
Despite the gaps in the data on private grants, the Institute for Higher Education Policy
conducted a survey to approximate the amounts given in the 2003-2004 academic year,
and determined that somewhere around $3 billion, with estimates as low as a $450
million to as high as $13Billion (see Figure 1 below). The most recent NCES survey
only had less than 600 respondents to the private grant question, of which they
interviewed a sample (Institute for Higher Education Policy, 2005). To date, there is no
comprehensive research on the number of private scholarships or the amount of the
awards and their impact on access or participation in higher education.
Institutional Responses to State and Federal Aid
The limited numbers of studies that have accounted for the interactions between
institutional, state and federal aid tend to focus on the institutional response to changes in
either state or federal policies. For example, Acosta (2001) examined the response of
colleges to changes in federal financial aid. She found that private institutions increased
the amounts of their tuition and institutional aid to students, whereas public institutions
increased tuition revenues and decreased student aid (Acosta, 2001). Similarly, Long
18
(2003) found that comparable institutional strategies to capture revenues were employed
in Georgia with the implementation of the HOPE Scholarship.
Figure 1. Estimated Student Aid by Source for Academic Year 2003-04
Institutional Aid ($23B)
State Aid ($6B)
Nonfederal Loans ($ 1IB)
Education Tax Benefits ($6B)
Private Aid ($3B)
Federal Pell Grants ($13B)
Federal Loans ($56B)
Other Federal Programs ($4B)
Federal Campus-Based ($3B)
Source: Institute for Higher Education Policy Survey, 2004-05; College Board, 2004
Research on the interaction between university policies and state merit aid
policies has shown that institutions utilize awards to attract students with high academic
achievement rather than closing the gap in need based financial aid left by statewide
merit aid programs (Doyle, Delaney & Naughton, 2004). The implications of these
findings for policy makers and administrators are significant. Both need to be fully
aware of the impact of their policies on the entire student population and the compound
affect it can have on students of low SES, who are most in need of financial aid.
19
State Merit Based Student Financial Aid
There are two main ways that states provide financial aid to students: through
loans and through need and merit based grants. The impact of state aid is particularly
significant when examining students with low SES (Alexander, 2001) or those of African
American and Hispanic origins (Heller, 1997). All 17 of the states that have adopted
broad-based merit scholarships have at least one of the following goals for its program: 1.
promote college access and attainment, 2. encourage, and/or reward students for working
hard, 3. reduce brain drain in the state by encouraging students to obtain their degrees
in the state where they live (Heller, 2002). (See Appendix A for details on each of the
states awards).
Until the late 1980s, only a small proportion of state appropriations were allocated
for merit-based funding to students (Cornwell, Mustard & Sridhar, 2003). The first merit
based financial aid program was the Arkansas Academic Challenge Scholarship,
introduced in 1991 by then Governor, Bill Clinton. However, since its inception in the
early 1990s, Georgias Helping Outstanding Pupils Educationally (HOPE) Scholarship
has led the way to increasing state supported broad-based merit scholarships.1 According
to a survey conducted by the Education Commission of the States, as of 2005, there were
17 states that had merit based scholarships similar to the one established by Georgia (see
Appendix A for complete program descriptions). The transition from funding need-based
scholarships to merit-based has been a relatively swift one, as it proves to be a politically
popular tool used in election bids for legislators in this era of growing accountability
1Georgia has both a HOPE Scholarship and a HOPE Grant. The HOPE Grant applies only to
non-degree programs and does not have a GPA requirement. Since the incentives that apply to
merit aid do not apply to the HOPE Grant, this study does not examine that component of the
Georgia Scholarship system.
20
(Dynarski, 2003). In 2000, the twelve states (at that time) with merit-based aid awarded
over $863 million in merit based aid and $308 million in need based aid in the 2000-01
academic year (Heller, 2002).
The combination of eligibility requirements, source of funding, and award
amounts in each state are the largest factors in determining the impact of each scholarship
program. In a state where there is limited funding, it makes sense to have more stringent
restrictions on who is awarded, in order to make the money last longer. A state might
also reduce the amount of funding per student in order to stretch tight dollars. In states
where the revenues are tied to state funding, such as in Louisiana and South Carolina,
there is a larger impact on funding during years where state fiscal budgets are tight
(Dynarski, 2002). States where lottery proceeds are the funding source, there is a
disproportionate amount paid in to the fund and very little returned to low socioeconomic
sectors of the state, which also generally tends to over-represent minorities (Binder, et.al.,
2003; Cornwell & Mustard, 2003b). Arkansas has had to limit new enrollees and West
Virginia had to cut need-based aid because it was based on the states budget whereas the
merit-based aid was based on lottery proceeds (Dynarski, 2003). A recent visit to the
website for the Washington merit scholarship program revealed that the Washington
Promise Scholarship program was terminated on June 30, 2006 (Washington Higher
Education Coordinating Board, 2006).
Eligibility for the wide reaching merit based state scholarships vary widely from
state to state. In Michigan, eligibility is solely based on a standardized test that is given
in 11th grade. In Florida, Louisiana, Mississippi, and West Virginia, the criteria for award
is based on grade point average and standardized test scores. Wyoming has the strictest
21
criteria for awards with both a high academic achievement requirement and high need
criterion. As a result of their high eligibility requirements, Wyoming only awards
between four and six awards annually. In Georgia, Kentucky, Nevada, and New Mexico
recipients are awarded solely based on grade point averages. Finally, Alaska awards its
scholars based on class rank. Basing awards on class rank is significant because blacks
have lower average grades in high school, which means a smaller proportion will meet
HOPESacademic requirements: nationwide, among those members of the high school
class of 1992 intending to go to college, 21 percent of whites had a high school GPA of
3.5 or above, while only 4 percent of blacks had such higher grades (National Center for
Education Statistics, 1995) (Dynarski, 2002).
The effect of award amounts has varied. In the case of Georgia, where the award
is significant and includes not only tuition but student fees and books; the effect has been
a significant shift towards students staying in state (Cornwell & Mustard, 2003b).
Because Georgias awards are based on lottery revenues, the state could, until recently,
well afford to pay for tuition, fees and books for students. One study by Cornwell and
Mustard (2002) theorized that due to the high correlation between pre-college academic
achievement and family income, HOPE scholarship funds would be capitalized in other
ways besides tuition. They found that doubling any countys HOPE scholarship
recipients would, on average, lead to a two percent rise in the number of registered cars
(Cornwell & Mustard, 2002).
In states where funding for the scholarship is limited, like Michigan, (which also
uses tobacco settlement dollars), the total award is $2500. There has been little incentive
effect in this state (Heller & Rogers, 2003); perhaps this is due in part because it is a one
22
time, non-renewable award. The Nevada Millennium Scholarship, with its funding base
in the limited tobacco dollars, had to make adjustments in terms of requirements for
awards and individual allotments during the 2005 legislative session (Ackerman, 2005).
More recent legislative changes to the program cut funding to students taking remedial
courses.
The results concerning the first goal of promoting college access and attainment
have been mixed. Dynarski (2003) found that generally, the new merit-based scholarship
programs increase participation in states by five to seven percentage points. On the other
hand, she also found that in the case of Georgias HOPE Scholarship, there was a
widening of the participation gap between blacks and whites. She speculated it was due,
at least in part, to the original stringent criteria placed on students in the HOPE
Scholarship.
In New Mexico, where tuition prices are also low, Binder, Ganderton, arid
Hutchins (2003) found that the New Mexico Success Scholarship did not increase
participation in that state for in-state students, but did increase participation for Native
American students. They also found that the scholarship tended to disproportionately
award white affluent students. In Nevada, early data indicate that the Nevada high school
continuation rate for first-time, degree-seeking college students in the fall semester
immediately following graduation increased from 32.4% in 1992 to 44.7% in 2002
(Herzog, 2005). While these numbers are encouraging for the future of Nevada, it
remains below the 2002 national average of 56.6% for this age group (NCHEMS, 2005).
The final measure of impact of merit-based scholarships is that of brain drain.
While there is evidence that students tend to stay in their home state when incentives are
23
offered, such as in the case of Georgias HOPE Scholarship (Cornwell & Mustard, 2002)
there is little research into the long-term retention of these citizens. Heller and Rogers
(2003) found that Michigans $2500 incentive did little to provide encouragement for
students to study harder, achieve more, or to pursue higher education in the state. They
offer as a warning to policymakers that it is still not known how much of an incentive is
necessary to retain students or how to focus incentives towards those students that might
benefit from an incentive system of rewards.
While several studies focus on the impact of a specific merit aid policy, none have
examined the aggregate impact of these policies in the states where they have been
adopted. In addition, the operational definition of merit aid remains broad and indicates
nothing more than the general criteria for award; this study seeks to deconstruct merit aid
programs and to estimate the impact of three broad characterizations of merit aid on
participation.
Postsecondary Participation
The pathway to college can take many different routes. For some, it includes
years of planning, including taking college preparation coursework; for others, it is a
decision that they make just before the start of a semester. This study is specifically
interested in examining the patterns of enrollment across states and institutions as they
are affected by merit-based aid and is not concerned about the process that students go
through in deciding whether to go to college.
Throughout the literature, enrollment is referred to in many different ways. For
example, within an economic framework, enrollment is referred to as demand, as
24
represented by those choosing to enroll in postsecondary education (Buss, Parker, &
Rivenburg, 2004). Other literature equates college choice with opportunity and access,
especially with regards to the long term impacts of where one decides to go to college
and future career opportunities (Gladieux & Swail, 1999; Thomas, 2003). In her 2004
study on the impact of financial aid on college choice by racial groups, Kim (2004) found
that the college choices of African American and Hispanic students were not significantly
influenced by financial aid. However, Whites and Asian American students were more
likely to enroll in their first choice college, given grants or loans or some combination of
the two.
The impact of financial aid on college choice by sector has been examined in
several studies (Jackson, 1978; Tierney, 1980; Pema & Titus, 2004). In their study on
the impact of cost, quality, and enrollment demand, Buss, Parker, and Rivenburg (2004)
found that increasing financial aid had a large positive effect on enrollment for students
who receive that aid. However, if tuition also increased, student enrollment would
decrease. Students appear to look beyond a net cost number and consider tuition and
aid separately. Perhaps this reflects uncertainty about continuation of aid in future years,
whereas tuition is forever (p. 65). Jacksons 1978 study found that the mere offer of
financial aid, regardless of amount, weighed significantly on the college choice decision.
Pema and Titus (2004) found that, state need-based financial aid and institutional
financial aid promote student choice among different types o f colleges and universities.
State need-based financial aid programs with relatively large awards per member of the
traditional college-age population appear to be particularly effective at promoting
enrollment in private four-year colleges and universities in a state (p. 520).
25
It is largely assumed that states that offer financial aid programs might improve
their participation rates. Participation rates are the number of students in postsecondary
education divided by the total population (Martinez, 2004). By investigating the
relationship between merit aid and participation, it may be possible to predict how
participation rates would be affected in the future. There has been some work done on
projecting state participation rates. For example, in his study of postsecondary
participation and state policy, Martinez (2004) proposed that states set benchmarks for
improving their postsecondary participation rates based on the states with the highest
participation rates. He found that if states maintained their current participation rates, the
top five states with the largest projected gaps for 2015 were located in the Southeast. The
participation gap is defined as the difference between the current college participation
rate in the state and that of a benchmark state. There is growing concern that while the
number of students enrolling in postsecondary education is increasing due to population
growth, the percentage of the population enrolling in postsecondary education is not
(Ruppert, 2003). Interestingly, four of the five states with the largest projected
enrollment gaps currently have merit based student financial aid programs in place: West
Virginia, Kentucky, Tennessee, and Mississippi. It may be possible that the effect of the
merit aid programs in these states would alter the projections that Martinez offers.
There is an established body of literature that points to the relationship between
financial aid and student enrollment; in particular, who benefits? and where do they
go? The impact of merit aid policies has not been examined to determine how they
effect first time undergraduate enrollment on the state and student levels. Additionally,
there is little evidence as to the impact of merit aid on enrollment across the states with
26
similar policies to determine if the effect is similar. This study seeks to fill that gap in
the research by examining the impact of merit aid on enrollment across similar merit aid
policies at the state and sector levels.
27
CHAPTER 3
METHODOLOGY
Introduction
This chapter provides an overview of the analytical procedures used in this study
and contains the following sections: Introduction, research questions, independent
variable description, methods of analysis, and summary. First, an explanation of the
methods that were used in this study, the sources of error for this type of analysis, and
how the errors were controlled for are outlined. Then, the research questions and the
corresponding hypotheses are provided. Drawn from the hypotheses, each of the
dependent and independent variables utilized in the hypotheses are identified and defined.
Finally, a description of each of the analyses that was conducted for each research
questions is provided. Then, the four main types of impact found as a result of this type
of analysis are examined. Finally, the regression equations and assumptions are
presented along with data limitations.
Main Methods
The main methods of analysis that were utilized in this study are interrupted and
pooled time series regression analysis. In addition, all regressions were checked for
autocorrelation using the Durbin-Watson statistic. Interrupted time series is useful when
examining a large scale policy adoption, such as the merit aid programs across states.
28
According to StatSoft, (2005), there are two main purposes in time series analysis: 1)
identifying the nature of the phenomenon represented by the sequence of observations,
and (b) forecasting (predicting future values of the time series variable).
Interrupted time series requires a series of observations of equal spacing before
and after an event. The first step in interrupted time series is to graph the data to
determine if there are any trends in the data line. Then, based on the visual inspection of
the trend line, a regression model is built to determine the impact of the intervention or
interruption (Meier & Brudney, 2002). Pooled time series is when there are a series of
observations of equal spacing of equal units. For example, in this study, there were data
for 1995-2004 reported for the institutions in Michigan. The data across all institutions in
that state were pooled for each of every year and the multivariate regressions were run on
the pooled enrollment data.
The data on enrollment in this study were likely to have a positive slope over
time, as enrollments have generally been increasing in most states over the last 25 years
(Mortenson, 2001). A concern when conducting time series analysis is that of
autocorrelation (Ostrom, 1990). Autocorrelation is where data from one year are
correlated with the prior years. For example, enrollment at institutions and in states is
generally correlated from one year to the next simply because there are issues of
population growth and capacity at work. In order to test for autocorrelation, the Durbin-
Watson statistic for each regression was calculated. In order to reduce some of the
autocorrelation, this study only examined the enrollment of first time undergraduates. By
excluding sophomores, juniors, and seniors from the data set, the autocorrelation that
corresponds with subsequent enrollments is reduced.
29
Sources o f Error
McDowall, McCleary, Meidinger, and Hay (1980) point out that there are three
main sources of error in time series: 1) trend; 2) seasonality; and 3) random error. Trend
is where a time series gradually drifts upward throughout most of its history
(McDowall, et al., p.l 5). Seasonality is when a series spikes consistently at a given
interval of time. Finally, random error is when a time series was detrended and
deseasonalized, observations would still fluctuate randomly about some mean level
(McDowall, et al., p i 4). For this analysis, a time variable (T) was included to account for
the trend component in this time series.
Pooled Time Series Analysis
For the pooled time series analysis, the impact of the type of merit aid program
(full, partial, or one-time payment) across states with similar programs was examined.
There were three intervention related dummy variables that were created for this study.
The first is a time trend (T), which accounts for the increasing trend in enrollment over
time, regardless of the policy adoption. The second is an intervention variable (TT)
which examines the immediate impact of the policy adoption. A significant TT indicates
that there was an immediate significant impact of the policy on enrollment. The third is a
longitudinal trend counter (TTT) which examines the long term impact of the policy
adoption. A significant TTT indicates that there was a significant long term impact of the
policy adoption on enrollment.
A dummy variable for each of the states by program types (1= Yes, 0=No for
each, One time payment, Partial tuition payment, Full tuition payment) was created.
30
Once the states have been assigned a code based on their merit aid program type, the data
were aggregated into the three large program categories. Then the data across each
condition separately were analyzed. The use of a dummy variable is recommended by
the Least Squares Dummy Variable (LSDV) model, as suggested by Sayrs (1989) for
studies with non-constant variation which is unique to each cross-section within the time
series. There are legitimate concerns regarding the aggregation of data over units of
analysis that are not comparable. However, the pooled design will quickly reveal
noncomparability because the disturbance vector will not fit a set of realistic assumptions
about the data (Sayrs, p. 16).
In his similar study of health reform policies across states, Stream (1999) used a
pooled cross-sectional time series data set that was constructed to examine factors related
to health reform across fifty states for a period of time to create a state-year variable.
In this study, a similar pooled variable was created for the enrollment in each state for a
given year, which served as the unit of analysis for research question 2, the state level
pooled time series analyses. A second pooled variable was created for the enrollment in
each sector in each state for a given year, which served as the unit of analysis for research
question 4, the sector level pooled time series analysis.
Levels of Analysis
The analysis in this study took place in two main stages, according to the research
questions. For the purposes of the dissertation, and in order to provide the reader with
clarity as to how the research questions relate to specific data sets and methodologies, the
31
reader is referred to Appendix A, which outlines the datasets and methodologies in a
matrix for ease of use.
State Level Analysis
1. How has merit aid impacted postsecondary participation in each of the states that
has adopted this student financial aid policy?
2. What is the effect of the following three types of merit aid programs on state-level
participation:
a) Full-tuition programs
b) Partial tuition programs
c) One time payment programs
Sector Level Analysis
3. How much has merit aid impacted participation by sector and level in the states
that have adopted this financial aid policy?
4. What is the effect of the following three types of merit aid programs on
enrollment across similar programs by sector and level?
a. Full tuition
i. 2-year public
ii. 2-year private
iii. 4-year public
iv. 4-year private
b. Partial tuition
i. 2-year public
ii. 2-year private
32
iii. 4-year public
iv. 4-year private
c. One time payment
i. 2-year public
ii.
2-year private
iii. 4-year public
iv. 4-year private
Hypotheses
Hypotheses were written for the research questions to allow for a test of
significant differences to a) compare participation before and after policy enactment of
merit policies, and b) compare participation differences across the three different types of
merit aid programs. The hypotheses for each of the questions are as follows:
1. How has merit aid impacted first time undergraduate postsecondary participation
in each of the states that has adopted this student financial aid policy?
For each state,
i. Ho: bb - ba= 0, the policy adoption did not have a significant effect
on enrollment in the state.
ii. Ha: bb - ba ^ 0, the policy adoption did have a significant effect on
enrollment in the state.
bb is the slope of the line before policy adoption, which was calculated using
regression analysis for the time series for each state separately
33
ba is the slope of the line after the policy adoption, which was be calculated
using regression analysis for the line in the time series after the policy
adoption for each state.
enrollment across similar programs:
Using pooled data for Full tuition programs
Using pooled data for Partial tuition programs
ii. Ha: bb - ba # 0, the policy adoption did have a significant effect on
Using pooled data for One time payment programs
bb is the slope of the line before policy adoption for the pooled state level data
which were calculated by aggregating the institutional data for each state and
then compiling the states data into program types. For the full tuition
34
programs, N=6, for the Partial tuition programs, N=6 and for the One time
payment programs, N=1.
ba is the slope of the line after the policy adoption for the pooled state level
data calculated by aggregating the institutional data for each state and then
compiling the states data into program types.
3. How has merit aid impacted participation by sector and level in the states that
have adopted this financial aid policy?
Using pooled data for each state,
i. 2-year public
1. Ho: bb - ba= 0, the policy adoption did not have a
significant effect on enrollment in the state.
2. Ha: bb - ba f 0, the policy adoption did have a significant
effect on enrollment in the state.
ii. 2-year private
iii. 4-year public
iv. 4-year private
2. Ha: bb - ba # 0, the policy adoption did have a significant
bb is the slope of the line before policy adoption for the pooled
institutional level data which is calculated by aggregating the institutional
data for each sector and level.
ba is the slope of the line after the policy adoption for the pooled
institutional level data calculated by aggregating the institutional data for
each state by sector and level.
enrollment across similar programs by sector and level?
Full tuition
i. 2-year public
significant effect on enrollment.
2. Ha: bb - ba i- 0, the policy adoption did have a significant
effect on enrollment.
ii. 2-year private
1. Ho: bb - b a= 0, the policy adoption did not have a
36
2. Ha: bb ba # 0, the policy adoption did have a significant
iii. 4-year public
1. Ho: bb - b a= 0, the policy adoption did not have a
2. Ha: bb - b a ^ 0, the policy adoption did have a significant
iv. 4-year private
2. Ha: bb - ba ^ 0, the policy adoption did have a significant
Partial tuition
i. 2-year public
ii. 2-year private
2. Ha: bb - ba ^ 0, the policy adoption did not have a
37
iii. 4-year public
iv. 4-year private
2. Ha: bb - ba ^ 0
One time payment
i. 2-year public
significant effect on enrollment
2. Ha: bb - ba ^ 0, the policy adoption did not have a
significant effect on enrollment
ii. 2-year private
1. Ho: bb - ba= 0 the policy adoption did not have a
iii. 4-year public
38
effect on enrollment
iv. 4-year private
where bb is the slope of the line before policy adoption and ba is the slope
of the line after the policy adoption.
Description of Analyses
In the first part of this study, which addresses research questions 1 and 2, this
research examines the impact of merit aid policies on enrollment of postsecondary
students on a broad state level. The second part of this study, research questions 3 and 4,
utilize institutional level data, to facilitate analyses of the data by sector (private/public)
and level (2 and 4-year). The analysis examined comparisons within sectors and levels
for the same states, but it also examined comparisons by sector and level across states
with similar aid programs to see if similarities in enrollment changes occurred before and
after policy adoptions.
First, a pooled time series analysis o f enrollment for each o f the merit aid states
was separately conducted. Meyer (1995) argues that, government policies often create
natural treatment and comparison groups. Frequently, this event occurs because our
federal system of government allows one state to change a policy while others do not.
39
The many cross-state differences in policies and changes in these policies allow the
examination of a wide range of questions (p. 158)
Merit aid policies provide a unique opportunity for higher education scholars to
examine the impact of state level financial aid on participation at the state and
institutional levels. Specifically, there are three main types of merit aid that these
policies provide for examination: 1) One time payment, as in the case of Michigan; 2)
Partial payment, as in the case of Missouri, Mississippi, and 3) Full tuition payment, as in
the case of Georgia and Florida. For the second phase of analyses, the data were
aggregated across states with similar aid types in order to examine the differential effects
of each type of merit aid program.
The third step in this analysis examined how much merit aid impacted
participation by sector and level in the states that have adopted this financial aid policy.
This was done using time series analysis with institutional data that is aggregated by
sector (private and public) and level (2-year and 4-year) within each state.
The fourth analysis was a time series analysis to determine the impact on
enrollment by type of aid program in each sector and level across states with similar aid
programs. As described earlier, this by compiling the institutional data for each of the
states with similar aid types, plotting this data against time for each program type by
sector and level, and then calculating a regression line for each sector.
Prior research suggests that students will go to the higher cost institution when
their choice is on the margins (Dynarski, 2003). However, the magnitude of the impact
has yet to be determined as well as if that impact is uniform across all types of merit aid
programs. The analysis in this study provides regression equations that fit the data for the
40
observed trends in each of the states, merit aid program types, within sectors across states
or sectors within a single state.
Impact Analysis
According to McDowall, et. al. (1980), in time series impact analysis, there are
four main trends that emerge from the graphs (See Figure 2, below). The first pattern that
may emerge is the gradual and permanent change in the trend. For this study, the change
would be indicated by gradual change in enrollment after the adoption of a merit aid
policy; that is, the slope and the intercept of the line changes gradually over time. The
second is a gradual and temporary impact, which is indicated by a slow rise in the data
line then a decline back to the pre-intervention state. The intercept changes and slope
both change and then return back to their original state. The third impact is an abrupt
initial impact and a permanent long term change, as indicated by a change in both the
slope of the line as well as the intercept that remains over time. The fourth and final
impact is an abrupt initial impact with a gradual return to the initial state.
If legislators are looking to merit aid to make a significant long-term impact on
enrollment, they are likely going to want to see a permanent impact and most would
probably prefer an abrupt change as well. The slope of the line is significant because it
indicates the rate of participation in this study. As outlined earlier, Martinez (2004)
determined the participation gaps across the states. The four o f the five states with the
most significant gaps in participation are merit aid states. If merit aid proves to be a
useful tool in closing this gap, it could prove to be a successful strategy for other states
41
with significant participation gaps to utilize in an effort to raise the level of postsecondary
participation in their states.
Figure 2. Impact Patterns
Duration
Permanent Temporary
15
3
y \
03
S-4
o
O
n
s
e
t
A
b
r
u
p
t
A----- * ---*---*---A
A---A---A
Source: McDowall, et. al. (1980)
Modeling Using Regression Analysis
The initial equations for the time series analysis were estimated to determine the
impact of the award on enrollments in each of the merit aid states. The initial equation
estimated for enrollment in each state (at both the state and sector/level of analysis) is:
Y t= bo + b i T + b 2T T + b s T T T + I ^ X h s g r a d + b5XHEAPP+ b6X[NSTTUiT + b7XuNEMP + ^
42
where Yt = number of time-series observations for enrollment. T= a time counter from 1
to N, where N is the number of years observed for each state; so some states will have 7
years while others have 12, depending on how long ago the policy was adopted. TT = a
dummy variable coded 0 for those years before the adoption of merit aid policy and 1 for
the years after. This dummy variable is an indicator of when the state adopted the policy.
TTT = a dummy variable coded 0 for the years before adoption and 1, 2, 3, .. .for years
after the adoption of the policy and serves as a counter for the number of years in which
the policy is in effect; and et= the error term (Bingham & Felbinger, 2002).
Additionally, HSGRAD = the number of public high school graduates in the state.
HEAPP = the higher education appropriations in the state. INSTTUIT = average in state
tuition. UNEMP = annual unemployment rate in the state. The independent control
variables were selected based on previous work by Long (2003), where she investigated
the impact of the HOPE merit aid policy adoption on these factors.
The assumptions for regression analysis apply to the use of regressions in time
series analysis. The basic assumptions for regressions according to Lewis-Beck (1980)
are: 1) No specification error, that is that the relationship between Xj and Yj is linear and
that no relevant independent variables have been excluded and no irrelevant independent
variables have been included. 2) No measurement error, which is that the variables Xj and
Yi are accurately measured. 3) The following assumptions concern the error term, eL: zero
mean E (;) = 0. For each observation, the expected value of the error term is zero.
2 2
Homoscedasticity E(sj ) = 6 . The variance of the error term is constant for all values of
Xj. No autocorrelation: E(8,j) = 0 (i ^ j). The error terms are uncorrelated. The
43
independent variable is uncorrelated with the error terms E (eiXi) = 0. Normality. The
error term, sc, is normally distributed (p. 26).
Data
Prior to analysis, the researcher applied for permission to conduct research using
secondary data on human subjects by the University of Nevada, Las Vegas, Office for the
Protection of Research Subjects. Permission to conduct this research was granted on
March 13, 2006. Exempt research review status was applied to this research because the
National Center for Education Statistics (NCES) Integrated Postsecondary Education
Data System (IPEDS) was the source for the dependent variable, first time undergraduate
student enrollment at an institutional level. The institutional participation for each of the
15 states (Alaska, Arkansas, Florida, Georgia, Kentucky, Louisiana, Michigan,
Mississippi, Missouri, New Mexico, Nevada, South Carolina, Tennessee, Washington,
and West Virginia) was downloaded for each year.
This study focuses on the analysis of the 12 states with more than five years of
data since the policy adoption. However, data from Tennessee, Washington, and West
Virginia were used also collected to provide an early indication of their enrollment
response. In the case where institutional data were missing, the institutional outliers were
noted and removed from the data set (see Appendix E for the removed institutions). For
state level participation, the IPEDS institutional data for each state were compiled in
order to calculate the state level participation rates for each year. For the sector level
participation, the IPEDS institutional data for each sector in each state were compiled in
order to calculate the sectors participation rates for each year. The data were limited by
44
selecting to examine the only the first time undergraduate population since merit aid
applies to undergraduate students and initial enrollment.
The independent variables in this study were tuition (INSTTUIT), state higher
education appropriations (HEAPP), number of public high school graduates (HSGRAD),
annual unemployment rate (UNEMP), and three dummy variables that were coded to
determine the impact of the merit aid policies on enrollment. The first dummy variable
(T) was a time counter (1,2,3,.. .n); the second dummy variable (TT) was an indicator for
when the program began, program = 0 before adoption, 1 after adoption; the third dummy
variable (TTT) was a trend indicator where trend = 0 before adoption and 1,2,3,.. .n after
adoption. The in-state undergraduate tuition data were downloaded at the same time as
the enrollment data from IPEDS for each year.
For research question RQ1, the pooled annual average in-state resident tuition was
calculated, which included two and four year private and public institutions. The state
higher education appropriations were downloaded from the Illinois State Universitys
Grapevine website for each year. The number of public high school graduates was
retrieved from the Pell Institute and Tom Mortensons Postsecondary OPPORTUNITY
website. The annual unemployment rate for each state was downloaded from the Bureau
of Labor Statistics (BLS) website.
There are three states that have adopted merit aid policies in the last three years:
Tennessee, Washington, and West Virginia. The enrollment in these states served as a
test for the regression equations developed out of the models for each of the merit aid
programs. Additionally, South Carolina had two programs that were adopted within five
years of each other; the latter was modeled after the Georgia HOPE and bears the same
45
name. The data for South Carolina was examined with both merit aid programs included
in the analysis. All of the states selected for this study have at least five years of data post
adoption of the merit aid policy. Additionally, programs that only serve 4-6 students, as
in the case of Wyoming, were excluded because of the limited impact on participation.
Summary
This chapter presented the four main research questions along with hypotheses.
Two of the research questions concerned the impact of the policy adoption at a state level
and two at the sector level. There were 12 states that were fully examined in this study.
An additional three states, West Virginia, Washington, and Tennessee, are briefly
examined for initial impact since their policies have been adopted in the last five years.
Then, outlined the methods of this study, which used pooled and time series
regression models to examine the impact of state merit aid scholarship programs on
postsecondary enrollment. It also provided a detailed account of the impact analysis and
subsequent analysis that took place in this study.
This chapter also presented a summary of all of the data that was used for this
study and identified the sources for both the independent and dependent variables.
Independent variables were selected from previous studies on merit aid and the dependent
variable was selected based on eligibility from the initial phase of the adoption.
46
CHAPTER 4
STATE LEVEL ANALYSIS RESULTS
Introduction
Chapters 4 and 5 present the results of the statistical analysis of the research
questions of this study. Chapter 4 contains the results of the state level analyses in
research question RQ1 and RQ2. Chapter 5 contains the results of the sector level
analyses in research questions RQ3 and RQ4. The purpose of this study was to examine
the differential impact of the comprehensive state-wide merit scholarship programs
across states, sectors and institutions. Both chapters 4 and 5, which present the results of
this study, are organized by research question. The first two research questions are
concerned with the impact of merit aid policies on a state level and are contained in
chapter 4. Research question one (RQ1), how has merit aid impacted non-profit
postsecondary participation in each of the states that has adopted this student financial aid
policy? Research question two (RQ2), what is the effect of the following three types of
merit aid programs on state-level participation: Full-tuition programs; Partial tuition
programs; and One time payment programs.
For research question RQ1, analyses of the determinants of state level enrollment
were conducted in each state using time series regression. Institutional enrollment data
was compiled for each year and used as the state level enrollment figure in the analysis.
Research question RQ2 was an analysis of the determinants of enrollment across states
47
with similar merit aid types using pooled time series analysis. This analysis was
conducted in three stages. First, enrollment data of states with full tuition programs was
pooled and analyzed, then states with partial tuition payment programs, and then the one
state with one time payment program was analyzed using backward multiple regression
analysis.
For all of the backward stepwise time series regression models, variables in the
model that produced the smallest change in the r2 and where the probability of F-to-
remove >. 1O no longer held were removed. The significant time series graphs and
backward multiple regression analysis results are presented in this chapter. The graphs in
this chapter are the time series for each state. The tables in this chapter include (a) the
multiple correlation coefficients, R, (b) the coefficient of determination r squared or r2,
(c) adjusted r2, (d) unstandardized coefficients B with standard error (Std.error), (e)
standardized coefficients Beta, t-statistic, and (f) significance (Sig.). Summaries of all of
the regression outputs for research question RQ1 can be found in Appendix B.
Finally, autocorrelation is a concern with time series analysis and therefore the
Durbin-Watson statistic is calculated in this study as a test of autocorrelation. Only those
Durbin-Watson statistics that were of concern because they were outside of the upper or
lower limits of the statistic (given the degrees of freedom and number of cases) and
demonstrated the existence of autocorrelation in the data are presented in this chapter. A
complete table of the Durbin-Watson statistic outputs can be found in Appendix E. The
analysis was performed using SPSS 13.0 for Windows.
48
Research Question 1
Research question 1: How has merit aid impacted first time undergraduate
policy?
For research question 1, each state is presented with the hypotheses, then a graph
of the data, then the entry and final regression equations, then a table of the analysis with
a summary description. For each state, the N = years of observations. For example, in
Alaska, there were ten years of pooled enrollment data so N=10.
Alaska (AK)
The first hypothesis states that the slope of the regression line before the policy
adoption was equal to the slope of the line after the intervention in each state. The
alternative hypothesis is that the slopes are not equal. These hypotheses are represented
by the equations:
Ho: bb - ba= 0, the policy adoption did not have a significant effect on enrollment in the
state.
Ha: bb - ba 4- 0, the policy adoption did have a significant effect on enrollment in the state
The entry and final regression models are:
Entry model: ENROLL = HEAPP + HSGRAD + INSTTUIT + T + TT + TTT + UNEMP
Final model: ENROLL = HSGRAD + TT + TTT +UNEMP + CONSTANT
Below is the graph of the time series for enrollment in Alaska, with an
intervention line indicating the year of the policy adoption.
Figure 3. Annual Enrollment Pre- and Post- Merit Aid Adoption, Alaska
49
3000
2700
2400
iU 2100
1800
1500
1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
Year
The significance of F(4, 9) = 78.759, p<.001 is well below the a< .05. The figure
below provides the unstandardized coefficients, standardized coefficients, and standard
error for the final model. More than 97 percent (adjusted R2 = .972) of the variation in
enrollment is explained by the number of high school graduates in the state, the short and
long term impact of the policy adoption and unemployment. Approximately 2 percent of
the variance is due to other factors other than those initial factors stepped out as not
significant in this state. The standard coefficients (beta) estimates relative predictive
power of unemployment, policy adoption, long term impact, and the number of high
50
school graduates. Since the short term impact (TT) and long term impact (TTT) is
significant, the H0 is rejected.
Table 1
Regression Analysis Coefficients, Final Model for Alaska3 (N=10)
Unstandardized
Coefficients
Standardized
Coefficients
t
Sig.
Final Model B Std. Error Beta
(Constant) 1421.056 860.275 1.652 .159
HSGRAD .408 .102 .632 4.004 .010
TTT 169.223 34.581 .932 4.894 .004
TT -476.621 95.716 -.728 -4.980 .004
UNEMP -277.457 51.799 -.507 -5.356 .003
Note: R= .992, R Square = .984, Adjusted R Square = .972.
3Dependent Variable: First time undergraduate enrollment (ENROLL), a = .05
Arkansas (AR)
adoption is equal to the slope of the line after the intervention in each state. The
by the equations:
state.
51
Ha: bb - ba ^ 0, the policy adoption did have a significant effect on enrollment in the
state.
Below is the graph of the time series for enrollment in Arkansas, with an
Figure 4. Annual Enrollment Pre- and Post- Merit Aid Adoption, Arkansas
20000-
19000-
c
0
E
18000-
O
i -
c
111
17000-
16000-
1986 1988 1990 1992 1994 1996 1998 2000 2002 2004
Year
Entry model: ENROLL=HEAPP + UNEMP + HSGRAD + INSTTUIT + T + TT + TTT
Final model: ENROLL = UNEMP + TTT + T + CONSTANT
52
The significance of F(3, 18) =16.786, p<.001 is well below the a< .05. The figure
error for the final model. Nearly 73 percent (adjusted R = .725) of the variation in
enrollment is explained by unemployment, time, and the long term impact of the policy
adoption. Approximately 20 percent of the variance is due to other factors other than
those initial factors stepped out as not significant in this state. The standard coefficients
(beta) estimates relative predictive power of higher education appropriations, time, the
long term impact of the policy adoption, and unemployment. Reject H0 since the long
term (TTT) impact of the policy was significant.
Table 2
Regression Analysis Coefficients, Final Model for Arkansas3(N=19)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant) 7316.377 1802.698 4.059 .001
UNEMP 877.156 188.636 1.138 4.650 .000
T 1105.536 162.658 6.195 6.797 .000
TTT -1019.994 163.838 -4.991 -6.226 .000
53
Florida (FU
by the equations:
state.
state.
Entry model: ENROLL= HEAPP + UNEMP + HSGRAD + INSTTUIT + T + TT + TTT
Final model: ENROLL = HEAPP + HSGRAD + INSTTUIT + T + TT + TTT +
CONSTANT
The significance of F(6, 12) = 122.291, p<.001 is well below the a< .05. Table 3
provides the unstandardized coefficients, standardized coefficients, and standard error for
the final model. More than 98 percent (adjusted R = . 984) of the variation in
enrollment is explained by the number of high school graduates, time, the adoption of the
merit aid policy and its long term impact, in state tuition, and higher education
appropriations. Approximately one percent of the variance is due to other factors other
than those initial factors stepped out as not significant in this state. The standard
coefficients (beta) estimates relative predictive power of higher education appropriations,
the number of high school graduates, time, the short and long term impact of the policy
54
adoption, and in state tuition. Reject H0 since the long (TTT) and short (TT) term impact
of the policy adoption was significant.
Below is the graph of the time series for enrollment in Florida, with an
Figure 5. Annual Enrollment Pre- and Post- Merit Aid Adoption, Florida
140000-
120000"
E 100000"
80000 -
60000
1992 1994 1996 1998 2000 2002 2004
Year
55
Table 3
Regression Analysis Coefficients, Final Model, for Florida3 (N=13)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant) 392939.30 55295.696 7.106 .000
HEAPP -4.12E-005 .000 -1.409 -5.058 .002
HSGRAD -1.855 .492 -1.579 -3.772 .009
TTT 19572.050 3138.997 3.326 6.235 .001
TT 12585.582 3876.971 .367 3.246 .018
T 7898.025 1662.336 1.773 4.751 .003
INSTTUIT -24.349 4.665 -1.564 -5.220 .002
Georgia (GA1
by the equations:
state.
56
state.
Below is the graph of the time series for enrollment in Georgia, with an
Figure 6. Annual Enrollment by Year Pre- and Post- Merit Aid Adoption, Georgia
c
0)
E
o
i -
c
LU
50000
1986 1988 1990 1992 1994 1996 1998 2000 2002 2004
Year
Final model: ENROLL = UNEMP + T + CONSTANT
The significance of F(2,l 8) = 37.621, p<.001 is well below the a< .05. The figure
57
2
enrollment is explained by unemployment and time. Approximately 17 percent of the
variance is due to other factors other than those initial factors stepped out as not
power of the number of time (T) and unemployment (UNEMP). Accept H0 since there
was no significant long (TTT) or short (TT) term impact of the policy adoption.
Table 4
Regression Analysis Coefficients, Final Model, for Georgia3(N=19)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant) 12573.466 9927.583 1.267 .223
T 1878.085 230.509 1.126 8.148 .000
UNEMP 5088.554 1661.784 .423 3.062 .007
Note: R= .908, R Square = .825, Adjusted R Square =.803.
Kentucky (KY)
by the equations:
58
state.
Ha: bb - ba 7^0, the policy adoption did have a significant effect on enrollment in the
state.
Below is the graph of the time series for enrollment in Kentucky, with an
Figure 7. Annual Enrollment by Year Pre- and Post- Merit Aid Adoption, Kentucky
35000
34000
33000
32000
31000
30000
1996 2002 2004 1994. 1998 2000
Year
Entry model: ENROLL- HEAPP + UNEMP + HSGRAD + INSTTUIT + T + TT + TTT
Final model: ENROLL = INSTTUIT + T + TTT + CONSTANT
59
error for the final model. More than 78 percent (adjusted R2 = .782) of the variation in
enrollment is explained by the long term impact of the policy adoption (TTT), time (T),
and the average in state tuition (INSTTUIT). Approximately 22 percent of the variance
is due to other factors other than those initial factors stepped out as not significant in this
state. The standard coefficients (beta) estimates relative predictive power of time (T), the
long term impact (TTT) of the policy adoption and in state tuition (INSTTUIT).
Table 5
Regression Analysis Coefficients, Final Model, for Kentucky3(N=l 1)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant) 59559.338 6639.310 8.971 .000
TTT 2477.449 632.461 4.112 3.917 .006
T 1888.717 391.005 4.607 4.830 .002
INSTTUIT -9.592 2.160 -7.981 -4.440 .003
60
Louisiana (LA)
by the equations:
Ho: bb - ba= 0, the policy adoption did not have a significant effect on enrollment.
Ha: bb - ba ^ 0, the policy adoption did have a significant effect on enrollment
Figure 8 is a graph of the time series for enrollment in Louisiana, with an
Figure 8. Annual Enrollment by Year Pre- and Post- Merit Aid Adoption, Louisiana
44000 -
42000 -
c
0)
E
o
V-
40000
c
LU
38000 -
36000
1993 1995 1997 1999 2001 2003
61 Year
Final model: ENROLL = HSGRAD + T + TTT + CONSTANT
Table 6
Regression Analysis Coefficients, Final Model, for Louisiana3(N=12)
Unstandardized
Coefficients
Standardized
Coefficients
t
Sig.
(Constant) -30150.46 18288.012 -1.649 .138
HSGRAD 2.031 .554 1.246 3.663 .006
TTT 1862.996 729.625 2.034 2.553 .034
T -1382.392 633.329 -2.088 -2.183 .061
The significance of F(3, 11) = 10.576, p<.001 is well below the a< .05. The
figure below provides the unstandardized coefficients, standardized coefficients, and
standard error for the final model. Almost 74 percent (adjusted R2 = .740) of the
variation in enrollment is explained by the number of public high school graduates in the
state (HSGRAD), the long term impact (TTT) of the policy adoption, and time (T).
Approximately 26 percent of the variance is due to other factors other than those initial
factors stepped out as not significant in this state. The standard coefficients (beta)
62
estimates relative predictive power of the number o f high school graduates (HSGRAD),
time (T), and the long term (TTT) impact of the policy adoption.
Michigan (Mil
by the equations:
Ha: bb - ba ^ 0, the policy adoption did have a significant effect on enrollment.
Since Michigan had only 10 cases and there were 7 independent variables that
were in the entry model, the Durbin-Watson statistic showed that there was negative
autocorrelation (DW= 3.577). In order to correct for this serial correlation, the researcher
reduced the number of independent variables to include only those that would examine
the impact of the policy adoption, time (T), program (TT), and long term impact (TTT).
This reduction of the number of independent variables brought the Durbin-Watson
statistic to within the acceptable range (DW=2.67). The entry and final regression
models are:
Entry model: ENROLL= T + TT + TTT
Final model: ENROLL = T + TT + TTT + CONSTANT
Figure 9 is the graph of the time series for enrollment in Michigan, with an
63
Figure 9: Annual Enrollment by Year Pre- and Post- Merit Aid Adoption, Michigan
90000 -
87000 -
84000 -
o 81000 -
78000
75000 -
1995 1997 1999 2003 2001
Year
The significance of F(3,9) = 32.194, p<.001 is well below the a< .05. The figure
error for the final model. More than 91 percent (adjusted R = .912) of the variation in
enrollment is explained by the short (TT) and long (TTT) term impact of the policy
adoption and time. Approximately 9 percent of the variance is due to other factors other
coefficients (beta) estimates relative predictive power of time and the short and long term
impact of the policy adoption. Reject H0 since there was a significant long (TTT) and
short (TT) term impact of the policy adoption on enrollment.
64
Table 7.
Regression Analysis Coefficients, Final Model, for Michigan3(N=10)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant) 72783.800 1560.781 46.633 .000
TT 4554.400 1940.305 .478 2.347 .057
T 1673.000 470.593 1.008 3.555 .012
TTT -1426.400 665.519 -.543 -2.158 .074
Missouri (MO)
by the equations:
state.
state.
Below is the graph of the time series for enrollment in Missouri, with an
65
Figure 10. Annual Enrollment by Year Pre- and Post- Merit Aid Adoption, Missouri
46000
44000 -
42000
c
a)
E
40000 -
o
L-
c
uu
38000 -
36000 -
1994 1996 1992 1998 2000 2002 2004
Year
Final model: ENROLL = TTT + CONSTANT
The significance of F(l,12) = 166.096, p<.001 is well below the a< .05. The
standard error for the final model. More than 93 percent (adjusted R2 = .932) of the
variation in enrollment is explained by the long term impact of the policy adoption.
66
estimates relative predictive power o f the long term impact (TTT) o f the policy adoption.
Reject H0 since there was a long term impact of the policy adoption.
Table 8
Regression Analysis Coefficients, Final Model, for Missouri3(N=13)
Unstandardized
Coefficients
Standardized
Coefficients
t
Sig.
(Constant)
TTT
38076.549 314.415
1022.913 79.371 .968
121.103
12.888
.000
.000
Mississippi (MS)
by the equations:
state.
state.
67
Below is the graph of the time series for enrollment in Mississippi, with an
intervention line indicating the year o f the policy adoption.
Figure 11. Annual Enrollment by Year Pre- and Post- Merit Aid Adoption, Mississippi
37500 -
35000 -
32500 -
C 30000 -
27500 -
25000 ~
1995 1997 1991 1993 1999 2001 2003
Year
error for the final model. Nearly 83 percent (adjusted R2 = .827) of the variation in
68
enrollment is explained by time and unemployment. Approximately 17 percent of the
variance is due to other factors other than those initial factors stepped out as not
power of time (T) and unemployment (UNEMP). Accept H0, there was neither a long
(TTT) nor a short (TT) term impact on enrollment after the policy adoption.
Table 9
Regression Analysis Coefficients, Final Model, for Mississippi3(N=14)
Unstandardized Standardized t Sig.
Coefficients Coefficients
(Constant) 13750.600 3840.165 3.581 .004
T 904.071 120.195 1.134 7.522 .000
UNEMP 1339.683 489.083 .413 2.739 .019
New Mexico (NMf
by the equations:
69
state.
state.
Below is the graph of the time series for enrollment in New Mexico, with an
Figure 12. Annual Enrollment by Year Pre- and Post- Merit Aid Adoption, New Mexico
16000-
15000-
c
0)
E
o
i -
c
m 14000-
13000"
2004 1992 1994 1996 1998 2000 2002
Year
70
The significance of F(l,12) = 21.922, p<.001 is well below the a< .05. The figure
enrollment is explained by the long term (TTT) impact of the policy adoption.
estimates relative predictive power of the long term impact of the policy adoption. Reject
H0 since there was a significant long term (TTT) effect of the policy adoption on
enrollment.
Table 10
Regression Analysis Coefficients, Final Model, for New Mexico3(N=13)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant)
TTT
13610.248 206.906
244.549 52.231 .816
65.780
4.682
.000
.001
71
Nevada (NY)
by the equations:
state.
state.
Final model: ENROLL = UNEMP + HEAPP + TT + INSTTUIT + CONSTANT
above provides the unstandardized coefficients, standardized coefficients, and standard
error for the final model. Nearly 94 percent (R2 = .935) of the variation in enrollment is
explained by the unemployment rate in the state (UNEMP), higher education
appropriations (HEAPP), the adoption of the policy (TT), and the average in state tuition
(INSTTUIT). Approximately 6 percent of the variance is due to other factors other than
those initial factors stepped out as not significant in this state. The standard coefficients
(beta) estimates relative predictive power of the number of unemployment rates
(UNEMP), higher education appropriations (HEAPP), the policy adoption (TT), and
instate tuition (INSTTUIT). Reject H0 since there is a significant short term impact (TT)
of the policy adoption.
72
Figure 13 is the graph o f the time series for enrollment in Nevada, with an
Figure 13: Annual Enrollment by Year Pre- and Post- Merit Aid Adoption, Nevada
14000"
C
0)
E
10000
o
I
c
UJ
8000 -
6000
1999 2001 1995 1997 2003
Year
73
Table 11
Regression Analysis Coefficients, Final Model, for Nevada8(N=10)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant) 4645.729 2720.136 1.708 .148
HEAPP -1.36E-005 .000 -.846 -2.195 .080
TT -3295.486 912.725 -1.149 -3.611 .015
INSTTUIT 2.602 .582 2.550 4.475 .007
UNEMP -1696.064 365.433 -.597 -4.641 .006
a Dependent Variable: First time undergraduate enrollment (ENROLL), a = .05
South Carolina (SC)
Since South Carolina had two policies that were adopted, there are two analyses
for this state. The first analysis is of the first policy adoption in 1998 and the second
examines the impact of the second policy adoption in 2001. These hypotheses are
represented by the equations:
state.
state.
74
Figure 14 is the graph o f the time series for enrollment in South Carolina, with an
intervention line indicating the year o f the first policy adoption.
Figure 14. Annual Enrollment by Year Pre- and Post- Merit Aid Adoption, South
Carolina
40000
37500
35000 -
E 32500 -
30000 ~
27500
25000 -
1993 1995 1997 1999 2001 2003
Year
75
enrollment is explained by time (T) and unemployment (UNEMP). Approximately 7
percent of the variance is due to other factors other than those initial factors stepped out
as not significant in this state. The standard coefficients (beta) estimates relative
predictive power of time (T) and unemployment (UNEMP). Accept H0 since there was
not a significant short (TT) or long (TTT) term impact of the policy adoption on
enrollment.
Table 12
Regression Analysis Coefficients, Final Model, for South Carolina3(N=12)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant) 24137.218 1254.332 19.243 .000
T 816.957 73.633 .912 11.095 .000
UNEMP 772.613 214.761 .296 3.598 .006
Final model: ENROLL = HSGRAD + T + TTT + CONSTANT
76
error for the final model. More than 96 percent (R2 = .963) of the variation in enrollment
is explained by the number of public high school graduates in the state, the impact of the
policy adoption, and time. Almost 4 percent of the variance is due to other factors other
coefficients (beta) estimates relative predictive power of the number of high school
graduates (HSGRAD), time (T), and the immediate impact (TT) of the policy adoption.
The second policy adoption had a significant positive impact on enrollment, reject H0.
Overall, enrollment was also significantly increasing as time went on.
Figure 15 is a graph for the second merit aid policy adoption in South Carolina,
Figure 15 Annual Enrollment by Year Pre- and Post- Merit Aid Adoption, South Carolina
40000-
o
c
L U
27500"
1993 1995 1997 1999 2001 2003
Year
77
Table 13
Regression Analysis Coefficients, Final Model, for South Carolina3(N=12)
(Constant) 14612.997 7821.373 1.868 .099
HSGRAD .501 .259 .133 1.931 .090
TT 41.63.359 696.487 .635 5.978 .000
T 305.054 104.642 .341 2.915 .019
Tennessee (TNI
by the equations:
state.
Ha: bb - ba 0, the policy adoption did have a significant effect on enrollment in the
state.
Figure 16 is the graph of the time series for enrollment in Tennessee, with an
intervention line indicating the year of the policy adoption. One year of data post-policy
78
intervention are not a sufficient number of observations to run a time series regression;
however the graph of the data are provided to give the reader a general impression of the
impact of the policy on enrollment.
Figure 16. Annual Enrollment by Year Pre- and Post- Merit Aid Adoption, Tennessee
1 650-
E
o
c
LU
2000 2002 1998 2004
Year
Washington (WA1
by the equations:
state.
79
E
n
r
o
l
l
m
e
n
t
Ha: bb - ba^ 0, the policy adoption did have a significant effect on enrollment in the
state.
Figure 17 is the graph of the time series for enrollment in Washington with an
intervention line indicating the year of the policy adoption. Two years of data post
policy intervention are not a sufficient number of observations to run a time series
regression; however, the graph of the data are provided to give the reader a general
impression of the impact of the policy on enrollment.
Figure 17: Annual Enrollment by Year Pre- and Post- Merit Aid Adoption, Washington
100000"
80000
60000
40000
20000
1997 1998 1999 2000 2001 2002 2003 2004
Year
80
West Virginia (WV)
by the equations:
state.
Ha: bb - ba # 0, the policy adoption did have a significant effect on enrollment in the
state.
Figure 18. Annual Enrollment by Year Pre- and Post- Merit Aid Adoption, West Virginia
18000-
16000-
14000
c 12000 "
10000"
8000 "
1997 1998 1999 2000 2001 2002 2003 2004
Year
81
Figure 18 is the graph of the time series for enrollment in West Virginia, with an
intervention line indicating the year of the policy adoption. Two years of data post
policy intervention are not an acceptable number of observations to run a time series
regression; however the graph of the data are provided to give the reader a general
impression of the impact of the policy on enrollment.
Research Question 2
Research Question 2: What is the effect of the following three types of merit aid
programs on first time undergraduate enrollment across states (a) full tuition payment
programs, (b) partial tuition payment programs, (c) one time payment programs?
This analysis was conducted in three stages, first the data for all of the full tuition
payment scholarship states was compiled and analyzed, then the states with partial tuition
payment programs, and then the one state with a one time payment program was
analyzed. States with full tuition payment programs include: AK, FL, GA, LA, NM, and
WV. States with partial tuition payment programs include: AR, KY, MS, MO, NV, and
SC. Since TN, WA and WV adopted the policy less than five years ago; there are an
insufficient number of years available so they are not included in the analysis. The one
time payment state is MI. For this research question, the data for each year for the states
is pooled by merit aid payment program. For example, there are five states with full
tuition programs that have been in place for more than five years and have varying
numbers of years of data, depending on when the program started in that state. N= X
number of years of data in each state.
82
Full Tuition Merit Aid Programs
Ha: bb - ba 7^0, the policy adoption did have a significant effect on enrollment.
Entry model: ENROLL- T + TT + TTT+ HEAPP + HSGRAD + UNEMP + INSTTUIT
Final model: ENROLL = T + HEAPP + HSGRAD + INSTTUIT + CONSTANT
Table 14
Regression Analysis Coefficients, Final Model, for Full Tuition3(N=67)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant) 1614.554 1930.174 .836 .406
HSGRAD .796 .062 .818 12.877 .000
HEAPP 8.87E-006 .000 .211 2.965 .004
INSTTUIT -2.033 .563 -.081 -3.613 .011
T 385.271 173.682 .052 2.218 .030
The significance of F(4,66) = 835.273, p<.001 is below the a< .05. The figure
error for the final model. More than 98 percent (adjusted R2= .981) of the variation in
83
enrollment is explained by the time, unemployment in the states, in state tuition, and the
number of high school graduates in each state. Only 2 percent of the variance is due to
other factors other than the program intervention, which was stepped out as not
power of higher education appropriations (HEAPP), the number of high school graduates
(HSGRAD), time (T), all of which had a positive impact on enrollment. The in state
tuition (INSTTUIT) had a negative impact on enrollment in the full tuition states. Accept
H0 since there was no significant long (TTT) or short term (TT) impact of the policy
adoption on enrollment.
Partial Tuition Merit Aid Programs
Ha: bb - ba f 0, the policy adoption did have a significant effect on enrollment.
Entry model: ENROLL= T + TT + TTT+ HEAPP + HSGRAD + UNEMP + INSTTUIT
Final model: ENROLL = HEAPP + HSGRAD + TTT + UNEMP + CONSTANT
The significance of F (4, 78) = 117.280, p<.001 is below the a< .05. The figure
error for the final model. Approximately 86 percent (adjusted R = .856) of the variation
in enrollment is explained by the number of high school graduates (HSGRAD) in each
state, and the long term (TTT) impact of the merit aid program adoption, the higher
education appropriations (HEAPP), and unemployment (UNEMP). Nearly 14 percent of
the variance is due to other factors other than the variables included in the entry model
which were stepped out as not significant in these states. The standard coefficients (beta)
84
estimates relative predictive power of the number of high school graduates (HSGRAD),
the long term impact of the policy intervention (TTT), unemployment (UNEMP) and the
higher education appropriations (HEAPP), all of which had a positive impact on
enrollment. Reject H0 since there was a significant negative long term impact of the
policy adoption.
Table 15
Regression Analysis Coefficients, Final Model, for Partial Tuition States3(N=79)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant) -14539.6 3205.163 -4.536 .000
HEAPP 2.47E-005 .000 .553 9.013 .000
HSGRAD .507 .052 .550 9.769 .000
TTT -348.448 135.170 -.118 -2.578 .012
UNEMP 1972.274 419.142 .219 4.706 .000
85
One Time Payment Merit Aid Programs
Since Michigan was the only state with a One time payment program, it had only
10 years of data and there were seven independent variables in the entry model, the
Durbin-Watson statistic showed that there was negative autocorrelation (DW= 3.577). In
order to correct for this serial correlation, the researcher reduced the number of
independent variables to include only those that would examine the impact of the policy
adoption, time (T), program (TT), and long term impact (TTT). This reduction of the
number of independent variables brought the Durbin-Watson statistic to within the
acceptable range (DW=2.67).
Entry model: ENROLL= T + TT + TTT
explained by time (T), and the short (TT) and long (TTT) term impact of the policy
adoption. Approximately 9 percent of the variance is due to other factors other than those
initial factors stepped out as not significant in this state. The standard coefficients (beta)
estimates relative predictive power of time, and the short and long term impact of the
policy adoption. Accept H0 since there was not a significant long (TTT) or short term
(TT) impact of the policy adoption on enrollment.
86
Table 16
Regression Analysis Coefficients, Final Model, for One time Payment State a (N=10)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant) 72783.800 1560.781 46.633 .000
TT 4554.400 1940.305 .478 2.347 .057
T 1673.000 470.593 1.008 3.555 .012
TTT -1426.400 665.519 -.543 -2.158 .074
Summary
Chapter 4 presented the results of the time series graphs and the time series
regression analysis of the data. In all, the longitudinal data for fifteen states were
graphed to determine the onset and duration of the impact the policy adoption had. The
analysis consisted of a backward multiple linear regressions to identify the predictive
value and relationship between the merit aid policy adoption and first time undergraduate
enrollment in the 15 states that have adopted these policies and across the three types of
merit aid programs.
Based on Longs (2003) work on the impact of the Georgia HOPE Scholarship on
institutions and tuition, three variables were identified as having significant impact: state
higher education appropriations, unemployment, and average tuition for in-state students.
87
Because the number of high school graduates in a state is directly linked to the number of
potential candidates for undergraduates, this study also accounts for the number of public
high school graduates in each state.
In the analysis of the impact the adoption of the merit aid policies in each state,
there was a significant positive short term impact on enrollment in Alaska, Florida, and
Nevada. In states with less than five years of data to analyze, there was a significant
short term impact on enrollment in West Virginia. There was a significant long term
impact of the policy adoption in eight of the 15 states that adopted these policies, Alaska,
Arkansas, Florida, Kentucky, Louisiana, Missouri, New Mexico, and Tennessee. For a
summary of these results, see Appendix D.
In addressing research question 2, across the five states with lull tuition payment
programs (AK, FL, GA, LA, and NM,), there was no significant long or short term
impact of the policy adoption on enrollment. Across the six states with partial tuition
payment programs (AR, KY, MS, MO, NV, SC), there was a statistically significant
negative impact of policy adoption and enrollment, as demonstrated in Figure 35. In the
state with a one-time payment program, there was no statistically significant impact of
the policy adoption on first time undergraduate enrollment.
88
CHAPTER 5
SECTOR LEVEL ANALYSES RESULTS
Introduction
The third and fourth research questions are concerned with the impact of merit aid
policies on the public 2 year, public 4 year, private 2 year and private 4 year institutions
in each of the states and across the three broad merit aid program types (full tuition,
partial tuition, and one time payment) and are presented in chapter 5. Research question
three (RQ3), how much has merit aid impacted participation by sector and level in the
states that have adopted this financial aid policy? And Research question four (RQ4),
what is the effect of the following three types of merit aid programs on enrollment across
similar programs by sector and level: full tuition; partial tuition; and one time payment?
An analysis of the determinants of sector and level enrollment (two year public,
four year public, two year private and four year private) was conducted for each state
using time series backward regression for research question RQ3. Finally, the analysis
for research question RQ4 consisted of a time series regression of the determinants of
sector and level enrollment across states with similar merit aid programs.
For all of the backward stepwise time series regression models, variables in the
model that produced the smallest change in the r2 and where the probability of F-to-
remove >.10 no longer held were removed. The significant time series graphs and
backward multiple regression analysis results are presented in this chapter. The graphs in
89
this chapter are the time series for each state. The tables in this chapter include (a) the
multiple correlation coefficients, R, (b) the coefficient of determination r squared or r2,
(c) adjusted r2, (d) unstandardized coefficients B with standard error (Std.error), (e)
standardized coefficients Beta, t-statistic, and (f) significance (Sig.). Summaries of all of
the regression outputs for research question RQ3 can be found in Appendix C.
Finally, autocorrelation is a concern with time series analysis and therefore the
Durbin-Watson statistic is calculated in this study as a test of autocorrelation. Only those
Durbin-Watson statistics that were of concern because they were outside of the upper or
lower limits of the statistic (given the degrees of freedom and number of cases) and
demonstrated the existence of autocorrelation in the data are presented in this chapter.
The analysis was performed using SPSS 13.0 for Windows.
Research Question 3
Research Question 3: How has merit aid impacted participation by sector and level in the
states that have adopted this financial aid policy?
For the analysis of the data in this research question, the N = a data point for each
institution in that sector for each year (five years before adoption and every year
thereafter) in that state. For each states sectors, N=(years * institutions) For example,
in Alaskas 2-year public sector, there was only one institution that reported data for the
required amount of time; so, N=years (10) * institutions (1) =10, which means that there
were ten years worth of data for one institution in that state.
Since this research question pertains to the sector and level of each state, the
results are presented by state and then separated into sector and level (i.e. 2 year public, 4
90
year public, 2 year private, 4 year private). States that did not have any private 2 year
institutions do not have that section reported here.
Alaska. 2-Year Public
Table 17
Regression Analysis Coefficients, Final Model, for Alaska, 2 year public a (N=10)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant) -3.90 2.718 -1.435 .201
T 3.900 .820 1.212 4.759 .003
TT 10.800 3.379 .584 3.196 .019
TTT -4.700 1.159 -.916 -4.055 .007
91
enrollment is explained by the short (TT) and long (TTT) term impact of the policy
adoption and time (T). Almost 7 percent of the variance is due to other factors other than
the variables included in the entry model which were stepped out as not significant in this
state. The standard coefficients (beta) estimates relative predictive power of the number
of time (T), and the short (TT) and long (TTT) term impact of the policy adoption.
Enrollment at two year public institutions immediately increased then over the long run,
decreased as a result of the long term impact of the policy adoption. Reject H0 since
there was both a long (TTT) and short term (TT) impact on enrollment.
Alaska. 4-Year Public and 2-Year Private
Alaska, 4-year public and 2-year private sector analysis yielded no significant
models from the analyses.
Alaska. 4-Year Private
Final model: ENROLL = T + HSGRAD + INSTTUIT + CONSTANT
The significance of F (3,19) = 3.583, p<.037 is below the a< .05. The figure
error for the final model. Nearly 29 percent (adjusted R - .290) of the variation in
enrollment is explained by in-state tuition (INSTTUIT), time (T) and the number of high
school graduates (HSGRAD). Almost 71 percent of the variance is due to other factors
92
other than the variables included in the entry model which were stepped out as not
power of in-state tuition, the number of high school graduates, and time. Accept H0 since
there is no significant long (TTT) or short term impact of the policy adoption on
enrollment.
Table 18
Regression Analysis Coefficients, Final Model, for Alaska, 4 year privatea (N=20)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant) -277.820 165.538 -1.678 .113
HSGRAD .065 .029 1.892 2.228 .041
T -9.688 5.298 -1.592 -1.829 .086
INSTTUIT -.006 .003 -.466 -2.078 .054
Arkansas. 2-Year Public
93
Final model: ENROLL = T + INSTTUIT + CONSTANT
The significance of F (2,194) = 15.207, p<.001 is below the a< .05. The figure
enrollment is explained by in-state tuition (INSTTUIT) and time (T). Almost 87 percent
of the variance is due to other factors other than the variables included in the entry model
which were stepped out as not significant in this state. The standard coefficients (beta)
estimates relative predictive power of in-state tuition (INSTTUIT) and time (T). The
policy adoption and time had a negative impact on enrollment in the two year public
schools in Arkansas. Accept H0 since there is no significant long (TTT) or short (TT)
term impact of the policy adoption.
Table 19
Regression Analysis Coefficients, Final Model, for Arkansas, 2 year publica (N=195)
(Constant) 730.655 888.299 .823 .412
T -62.888 18.765 -1.063 -3.351 .001
INSTTUIT .615 .112 .659 5.488 .000
94
Arkansas, 4-Year Public
Final model: ENROLL = TTT + INSTTUIT + CONSTANT
Table 20
Regression Analysis Coefficients, Final Model, for Arkansas, 4 year publica (N=195)
(Constant) 613.650 113.170 5.422 .000
TTT -40.641 14.770 -.286 -2.752 .007
INSTTUIT .319 .071 .469 4.523 .000
explained by in-state tuition (INSTTUIT) and the long (TTT) term impact of the policy
adoption. Almost 90 percent of the variance is due to other factors other than the
95
variables included in the entry model which were stepped out as not significant in this
state. The standard coefficients (beta) estimates relative predictive power of in-state
tuition (INSTTUIT) and the long term (TTT) impact of the policy adoption. The long
term (TTT) policy adoption had a significant negative impact on enrollment in the four
year public schools in Arkansas, therefore, reject H0.
Arkansas. 2-Year Private
Final model: ENROLL = T + TTT + INSTTUIT + CONSTANT
explained by in-state tuition (INSTTUIT), the long term (TTT) impact of the policy
adoption, and time (T). Almost 73 percent of the variance is due to other factors other
than the variables included in the entry model which were stepped out as not significant
in this state. The standard coefficients (beta) estimates relative predictive power of in
state tuition (INSTTUIT), the long term (TTT) impact of the policy adoption, and time
(T). The long term (TTT) impact of the policy adoption was negative in the two year
private schools in Arkansas. Reject H0 since there was a significant long (TTT) term
impact of the policy adoption on enrolment.
96
Table 21
Regression Analysis Coefficients, Final Model, for Arkansas, 2-year privatea (N=60)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant) -35.482 27.775 -1.278 .207
T 19.308 6.611 2.093 2.921 .005
TTT -21.914 6.803 -2.236 -3.221 .002
INSTTUIT .008 .004 .332 2.310 .025
Note: R= .518, R Square - .268, Adjusted R Square = .229.
Arkansas. 4-Year Private
Final model: ENROLL = T + TTT + HEAPP + INSTTUIT + CONSTANT
error for the final model. Nearly 16 percent (R = .155) of the variation in enrollment is
explained by in-state tuition (INSTTUIT), higher education appropriations (HEAPP), the
long term (TTT) impact of the policy adoption, and time (T). Almost 84 percent of the
97
variance is due to other factors other than the variables included in the entry model which
were stepped out as not significant in this state. The standard coefficients (beta)
estimates relative predictive power of in-state tuition (INSTTUIT), higher education
appropriations (HEAPP), the long term (TTT) impact of the policy adoption, and time
(T). The long term impact (TTT) of the policy adoption was negative on the two year
private schools in Arkansas. Reject the H0 since there was a significant long term (TTT)
impact.
Table 22
Regression Analysis Coefficients, Final Model, for Arkansas, 4 year privatea (N=206)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant) 107.814 42.065 2.563 .011
HEAPP -6.45E-008 .000 -.140 -2.071 .040
T 14.916 6.354 .356 2.347 .020
TTT -24.763 7.963 -.520 -3.110 .002
INSTTUIT .030 .006 .450 5.115 .000
Florida. 2-Year Public
98
Ha: bb - ba^ 0, the policy adoption did have a significant effect on enrollment.
Final model: ENROLL = T + UNEMP + CONSTANT
Table 23
Regression Analysis Coefficients, Final Model, for Florida, 2 year publica (N=496)
Coefficients Coefficients '
(Constant) -25.790 556.401 -.046 .963
UNEMP 157.237 76.315 .126 2.060 .040
T 92.557 25.548 .221 3.623 .000
The significance of F (2, 496) =6.731, p<.001 is below the a< .05. The figure
error for the final model. Only about 2 percent (adjusted R2 = .023) of the variation in
enrollment is explained by unemployment (UNEMP), and time (T). Nearly 98 percent of
the variance is due to other factors other than the variables included in the entry model
which were stepped out as not significant in this state. The standard coefficients (beta)
estimates relative predictive power of unemployment (UNEMP), and time (T), both of
99
which had a positive effect on enrollment. Accept H0 since there was no significant long
(TTT) or short (TT) term impact on enrollment.
Florida. 4-Year Public
Table 24
Regression Analysis Coefficients, Final Model, for Florida, 4 year public3(N=l 15)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant)
TTT
2146.475 219.894
250.677 55.033 .394
9.761
4.555
.000
.000
The significance of F (1, 114) = 20.748, pc.OOl is below the a< .05. The figure
enrollment is explained by the long term (TTT) impact of the policy adoption. Almost 85
100
percent of the variance is due to other factors other than the variables included in the
original entry model which were stepped out as not significant in this state. The standard
coefficients (beta) estimates relative predictive power the long term (TTT) impact of the
policy adoption. Reject H0 since there was a significant long (TTT) term impact on
enrollment.
Florida. 2-Year Private
There were no significant models derived from the analysis for the 2-year private
sector in Florida.
Florida. 4-Year Private
Final model: ENROLL - TTT + INSTTUIT + CONSTANT
The significance of F (2,519) = 128.583, p<.000 is well below the a< .05. The
standard error for the final model. Nearly 33 percent (adjusted R = .330) of the variation
in enrollment is explained by in state tuition (INSTTUIT) and the long term (TTT)
impact of the policy adoption. Almost 67 percent of the variance is due to other factors
other than the variables included in the original entry model which were stepped out as
not significant in this state. The standard coefficients (beta) estimates relative predictive
power of in-state tuition (INSTTUIT) and the long term (TTT) impact of the policy
adoption. An increase in the average in-state tuition (INSTTUIT) was associated with an
101
increase in enrollment. However, the long term impact o f the policy adoption had a
negative effect on enrollment in 4-year private institutions in the state. Reject H0 since
there was a significant long term (TTT) impact on enrollment.
Table 25
Regression Analysis Coefficients, Final Model, for Florida, 4 year private3(N=520)
(Constant) -35.654 26.803 -1.330 .184
TTT -14.520 4.801 -.116 -3.024 .003
INSTTUIT .040 .003 .606 15.826 .000
3Dependent Variable: First time undergraduate enrollment (ENROLL) a = .05
Georgia. 2-Year Public
Ha: bb - ba / 0, the policy adoption did have a significant effect on enrollment.
Final model: ENROLL = HSGRAD + INSTTUIT + UNEMP + CONSTANT
102
Table 26
Regression Analysis Coefficients, Final Model, for Georgia, 2 year public3(N=668)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant) -1263.526 327.105 -3.863 .000
UNEMP 80.322 25.533 .123 3.146 .002
HSGRAD .018 .005 .145 3.709 .000
INSTTUIT .413 .068 .242 6.049 .000
standard error for the final model. Nearly 8 percent (adjusted R2 = .086) of the variation
in enrollment is explained by in-state tuition (INSTTUIT), unemployment (UNEMP), and
the number of high school graduates (HSGRAD). Almost 92 percent of the variance is
due to other factors other than the variables included in the original entry model which
estimates relative predictive power of in state tuition (INSTTUIT), the number of high
school graduates (HSGRAD), and unemployment (UNEMP). An increase in the average
in-state tuition (INSTTUIT), unemployment (UNEMP), and the number of high school
103
graduates (HSGRAD) were all associated with an increase in enrollment. Accept H0
since there was not a significant long (TTT) or short (TT) term impact on enrollment.
Georgia. 4-Year Public
Final model: ENROLL = INSTTUIT + T + CONSTANT
Table 27
Regression Analysis Coefficients, Final Model, for Georgia, 4-year public3(N=668)
(Constant) -738.655 219.529 -3.365 .001
T -50.186 12.362 -.272 -4.060 .000
INSTTUIT 1.279 .146 .588 8.781 .000
104
in enrollment is explained by in-state tuition (INSTTUIT) and time (T). Almost 80
entry model which was stepped out as not significant in this state. The standard
coefficients (beta) estimates relative predictive power of average in-state tuition
(INSTTUIT) and time (T). An increase in the average in-state tuition (INSTTUIT) was
associated with an increase in enrollment. However, the enrollment trend was decreasing
over time in 4-year public institutions. Accept H0 since there was not a significant long
(TTT) or short (TT) term impact on enrollment.
Georgia, 2-Year Private
There was not a significant model that could be determined for the 2-year private sector
in Georgia because the N was too small.
Georgia, 4-Year Private
Final model: ENROLL = HEAPP + INSTTUIT + T + CONSTANT
The significance of F (3,402) = 28.931, p<.001 is well below the a< .05
threshold. The figure below provides the unstandardized coefficients, standardized
coefficients, and standard error for the final model. Nearly 17 percent (adjusted R2 =
.172) of the variation in enrollment is explained by in-state tuition (INSTTUIT), higher
education appropriations (HEAPP), and time (T). Almost 83 percent of the variance is
due to other factors other than the variables included in the original entry model which
105
estimates relative predictive power of in state tuition (INSTTUIT), higher education
appropriations (HEAPP), and time (T). Higher education appropriations (HEAPP) and
the average in-state tuition (INSTTUIT) had a significant positive effect on 4 year private
enrollment in Georgia. Accept H0 since there was not a significant long (TTT) or short
(TT) term impact on enrollment.
Table 28
Regression Analysis Coefficients, Final Model, for Georgia, 4 year private3(N=176)
(Constant) -113.899 141.481 -.805 .421
HEAPP 5.29E-007 .000 .504 2.326 .021
T -47.182 15.713 -.663 -3.003 .003
INSTTUIT .035 .004 .488 8.882 .000
Kentucky. 2-Year Public
106
Final model: ENROLL = INSTTUIT + UNEMP + T + CONSTANT
Table 29
Regression Analysis Coefficients, Final Model, for Kentucky, 2-year public3(N=270)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant) 1093.315 260.274 4.201 .000
UNEMP -126.955 52.164 -.186 -2.434 .016
T -35.194 13.422 -.289 -2.622 .009
INSTTUIT .208 .071 .363 2.943 .004
The significance of F (3, 269) = 3.157, p<.025 is below the a< .05 threshold. The
standard error for the final model. Nearly 2 percent (adjusted R2= .023) of the variation
in enrollment is explained by in-state tuition (INSTTUIT), unemployment (UNEMP), and
time (T). Almost 98 percent of the variance is due to other factors other than the
state. The standard coefficients (beta) estimates relative predictive power of in-state
tuition (INSTTUIT), unemployment (UNEMP), and time (T). Unemployment (UNEMP)
107
and time (T) both had a negative impact on 2-year public enrollment, while average in
state tuition (INSTTUIT) had a significant positive effect on enrollment. Accept H0 since
there was not a significant long (TTT) or short (TT) term impact on enrollment.
Kentucky. 4-Year Public
Final model: ENROLL = INSTTUIT + TTT + CONSTANT
standard error for the final model. Almost 28 percent (adjusted R2 = .281) of the
variation in enrollment is explained by in-state tuition (INSTTUIT) and the long (TTT)
term impact of the policy adoption. Nearly 72 percent of the variance is due to other
factors other than the variables included in the original entry model which were stepped
out as not significant in this state. The standard coefficients (beta) estimates relative
predictive power of in-state tuition (INSTTUIT) and the long (TTT) term impact of the
policy adoption. The long term (TTT) impact of the policy adoption had a significant
negative effect on enrollment for this sector. Finally, average in state tuition (INSTTUIT)
had a significant positive impact on enrollment at 4-year pubic institutions in Kentucky.
Reject H0 since there was a significant long term (TTT) impact on enrollment.
108
Table 30
Regression Analysis Coefficients, Final Model, for Kentucky, 4-year public3(N=88)
(Constant) -224.985 363.468 -.619 .538
TTT -184.323 57.671 -.480 -3.196 .002
INSTTUIT .933 .166 .845 5.621 .000
Kentucky 2-Year Private
There were not a sufficient number of institutions to conduct an analysis of this sector in
this state.
Kentucky. 4-Year Private
Final model: ENROLL = TTT + INSTTUIT + CONSTANT
The significance of F (2,236) = 22.059, p<.001 is below the a< .05 threshold.
The figure below provides the unstandardized coefficients, standardized coefficients, and
109
in enrollment is explained by in state tuition (INSTTUIT) and the long term (TTT)
power of in state tuition (INSTTUIT) and the long term (TTT) impact of the policy
adoption. Accept Hc since there was no significant long (TTT) or short term impact on
enrollment.
Table 31
Regression Analysis Coefficients, Final Model, for Kentucky, 4-year private3(N=237)
(Constant) 127.307 16.853 7.554 .000
TTT -6.755 4.014 -.108 -1.683 .094
INSTTUIT .012 .002 .423 6.606 .000
Louisiana. 2-Year Public
110
Entry model: ENROLL= HEAPP + HSGRAD + UNEMP + INSTTUIT + T + TT + TTT
Final model: ENROLL = HSGRAD + INSTTUIT + T + TTT + UNEMP + CONSTANT
Table 32
Regression Analysis Coefficients, Final Model, for Louisiana, 2-year public3(N=8)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant) 1801.022 781.623 -2.304 .022
HSGRAD .031 .018 .118 1.703 .089
INSTTUIT .781 .047 .600 16.663 .000
T 23.977 13.747 .230 1.744 .082
TTT -49.394 16.363 -.346 -3.019 .003
UNEMP 80.273 28.850 .186 2.782 .006
standard error for the final model. Almost 35 percent (adjusted R2 =.350) of the variation
in enrollment is explained by unemployment (UNEMP), the number of high school
graduates (HSGRAD), the long term (TTT) impact of the policy adoption, and time (T).
Nearly 65 percent of the variance is due to other factors other than the variables included
111
in the entry model which were stepped out as not significant in this state. The standard
coefficients (beta) estimates relative predictive power of unemployment, the number of
high school graduates (HSGRAD), the long term (TTT) impact of the policy adoption,
average in-state tuition (INSTTUIT), and time (T). The long term (TTT) impact of the
policy adoption on two year pubic enrollment was negative. However, the impact of the
number of high school graduates (HSGRAD), average in-state tuition (INSTTUIT), time
(T), and unemployment (UNEMP) all had a positive impact on 2-year public enrollment
in Louisiana. Reject H0 since there was a significant long term (TTT) impact of the
Louisiana. 4-Year Public
Final model: ENROLL - INSTTUIT + CONSTANT
Table 33 provides the unstandardized coefficients, standardized coefficients, and standard
t 'S
error for the final model. Nearly 15 percent (adjusted R =.151) of the variation in
enrollment is explained by in-state tuition (INSTTUIT). Almost 85 percent of the
estimates relative predictive power of in-state tuition (INSTTUIT), which had a positive
112
relationship to enrollment. Accept H0 since there was no long (TTT) or short (TT) term
impact of the policy adoption.
Table 33
Regression Analysis Coefficients, Final Model, for Louisiana, 4-year public3(N=156)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant)
INSTTUIT
-314.304 398.530
.921 .173 .395
-.789
5.340
.432
.000
Louisiana. 2-Year Private
There was significant missing data for in-state tuition (INSTTUIT) for the
institutions in this sector, which is why that variable was not included in the entry model.
Entry model: ENROLL= T + TT + TTT+ HEAPP + HSGRAD + UNEMP
Final model: ENROLL = HSGRAD + TT + CONSTANT
The significance of F (2,7) = 38.985, p<.001 is below the a< .05 threshold. Table
34 provides the unstandardized coefficients, standardized coefficients, and standard error
for the final model. Nearly 92 percent (adjusted R2 =.916) of the variation in enrollment
113
is explained by the number of high school graduates (HSGRAD) and the short term (TT)
power of the number of high school graduates (HSGRAD) and the short term (TT)
impact of the policy adoption. Reject H0 since there was a significant immediate (TT)
impact of the policy adoption on enrollment.
Table 34
Regression Analysis Coefficients, Final Model, for Louisiana, 2-year private3(N=8)
(Constant) 248.367 25.486 9.745 .000
HSGRAD -.006 .001 -1.271 -8.772 .000
TT 12.684 1.953 .941 6.493 .001
Louisiana. 4-Year Private
114
standard error for the final model. Nearly 65 percent (adjusted R =.649) of the variation
original entry model which were stepped out as not significant in this state. The standard
coefficients (beta) estimates relative predictive power of in-state tuition (INSTTUIT) and
time (T). Time (T) had a negative impact on 4-year private enrollment, while average in
state tuition (INSTTUIT) had a significant positive effect on enrollment. Accept H0 since
there was no significant impact for the long (TTT) or short (TT) term
Table 35
Regression Analysis Coefficients, Final Model, for Louisiana, 4-Year Private3(N=l 18)
(Constant) 14.292 57.471 .249 .804
INSTTUIT .060 .004 .842 14.664 .000
T -20.023 7.374 -.156 -2.716 .008
115
Michigan, 2-Year Public
Entry model: ENROLL= HEAPP + HSGRAD + INSTTUIT + T + TT + TTT + UNEMP
Table 36
Regression Analysis Coefficients, Final Model, for Michigan, 2-Year Public3(N=249)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant) -1864.295 400.342 -4.657 .000
HSGRAD .051 .008 .450 6.740 .000
UNEMP 66.326 25.843 .151 2.566 .011
INSTTUIT -.188 .033 -.376 -5.660 .000
enrollment is explained by average in-state tuition (INSTTUIT), unemployment
116
(UNEMP), and the number of high school graduates (HSGRAD). Almost 83 percent of
the variance is due to other factors other than the variables included in the original entry
model which were stepped out as not significant in this state. The standard coefficients
(beta) estimates relative predictive power of average in-state tuition (INSTTUIT),
unemployment (UNEMP), and the number of high school graduates (HSGRAD). The
impact of average in-state tuition (INSTTUIT) was negative and significant. The impact
of the number of high school graduates (HSGRAD) and unemployment (UNEMP) were
both positive and significant for 2-year public institutions. Accept H0 since there was no
significant long (TTT) or short (TT) term impact of the policy adoption on enrollment.
Michigan. 4-Year Public
standard error for the final model. Nearly 7 percent (adjusted R2=.072) of the variation
in enrollment is explained by average in-state tuition (INSTTUIT) and time (T). Almost
92 percent of the variance is due to other factors other than the variables included in the
entry model which were stepped out as not significant in this state. The standard
(INSTTUIT) and time (T). The impact of the time (T) on enrollment was negative but
117
not significant and the impact of average in-state tuition (INSTTUIT) was positive and
significant for 4-year institutions. Accept H0since there was no significant short (TT) or
long (TTT) term impact on enrollment.
Table 37
Regression Analysis Coefficients, Final Model, for Michigan, 4-Year Public3 (N=168)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant) 581.529 228.863 2.541 .012
T -46.305 24.739 -.176 -1.872 .063
INSTTUIT .326 .085 .361 3.829 .000
Michigan. 2-Year Private
Final model: ENROLL = HSGRAD + T + CONSTANT
The significance of F (2,64) = 3.690, p<.031 is below the a< .05 threshold. The
118
standard error for the final model. Nearly 8 percent (adjusted R2 =.078) of the variation
in enrollment is explained by time (T) and the number of high school graduates
(HSGRAD). Almost 82 percent of the variance is due to other factors other than the
state. The standard coefficients (beta) estimates relative predictive power of time (T) and
the number of high school graduates (HSGRAD). The impact of high school graduates
(HSGRAD) on enrollment was significantly negative while the impact of time was
positive and significant for 2-year private institutions. Accept H0 since there was no
significant long (TTT) or short (TT) term impact on enrollment.
Table 38
Regression Analysis Coefficients, Final Model, for Michigan, 2-Year Private3(N=65)
(Constant) 3066.463 1344.091 2.281 .026
HSGRAD -.066 .030 -1.033 -2.203 .031
T 73.684 29.051 1.189 2.536 .014
Michigan. 4-Year Private
119
Ha: bb - ba^ 0, the policy adoption did have a significant effect on enrollment.
Final model: ENROLL = HEAPP + INSTTUIT + UNEMP + CONSTANT
The significance of F (3, 443) = 56.247, p<.001 is below the a< .05 threshold.
in enrollment is explained by average in-state tuition (INSTTUIT), unemployment
(UNEMP), and higher education appropriations (HEAPP). Almost 73 percent of the
estimates relative predictive power of unemployment (UNEMP), average in-state tuition
(INSTTUIT), and higher education appropriations (HEAPP). The impact of the higher
education appropriations (HEAPP) and unemployment (UNEMP) on four year private
enrollment were significant and negative. However, the impact of the average in-state
tuition (INSTTUIT) was positive and significant. Accept H0 since there was no
120
Table 39
Regression Analysis Coefficients, Final Model, for Michigan, 4-Year Private3 (N=444)
Unstandardized Standardized t
Sig.
(Constant) 561.586 169.366 3.316 .001
HEAPP -4.44E-007 .000 -.231 -3.876 .000
UNEMP -55.877 20.026 -.162 -2.790 .005
INSTTUIT .037 .003 .555 12.987 .000
Missouri, 2-Year Public
Final model: ENROLL = INSTTUIT + CONSTANT
The significance of F (1,265) = 5.471, p=.020 is below the a< .05 threshold. The
standard error for the final model. Only about 2 percent (adjusted R2 =.017) of the
variation in enrollment is explained by average in-state tuition (INSTTUIT). Nearly 98
121
(INSTTUIT). The impact of average in-state tuition was positive and significant. Accept
H0 since there was no significant long (TTT) or short (TT) term impact on enrollment.
Table 40
Regression Analysis Coefficients, Final Model, for Missouri, 2-Year Public3(N=266)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant)
INSTTUIT
641.335 190.318
.175 .075 .142
3.370
2.339
.001
.020
Missouri. 4-Year Public
Final model: ENROLL = INSTTUIT + T + TTT + CONSTANT
122
in enrollment is explained by time (T), the long term (TTT) effects of the policy
adoption, and the average in-state tuition (INSTTUIT). Almost 87 percent of the
estimates relative predictive power of time (T), the long term (TTT) effects of the policy
adoption, and the average in-state tuition (INSTTUIT). The impact of time and average
in-state tuition (INSTTUIT) on enrollment was positive and significant. The long term
(TTT) impact of the policy adoption was negative and significant on enrollment in 4-year
public institutions in Missouri. Reject H0 since there was a significant long (TTT) term
impact on enrollment.
Table 41
Regression Analysis Coefficients, Final Model, for Missouri, 4-Year Public3(N=168)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant) -497.423 621.342 -.801 .425
T 252.126 120.878 .396 2.086 .039
TTT -394.831 194.738 -.394 -2.027 .044
INSTTUIT .448 .085 .405 5.259 .000
123
Missouri. 2-Year Private and 4-Year Private
In Missouri, the 2-year private and 4-year private did not yield models that were
significant for enrollment.
Mississippi. 2-Year Public
Final model: ENROLL = INSTTUIT + HEAPP + T + CONSTANT
standard error for the final model. Nearly 4 percent (adjusted R -.036) of the variation
in enrollment is explained by in-state tuition (INSTTUIT), higher education
appropriations (HEAPP), and time (T). Almost 96 percent of the variance is due to other
factors other than the variables included in the entry model which were stepped out as not
power of in-state tuition (INSTTUIT), higher education appropriations (HEAPP), and
time (T). Accept H0 since there was no significant short (TT) or long (TTT) term impact
on enrollment.
124
Table 42
Regression Analysis Coefficients, Final Model, for Mississippi, 2-year public3 (N=224)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant) 2281.263 720.246 3.167 .002
HEAPP -1.59E-006 .000 -.264 -1.735 .084
T 126.552 46.130 .532 2.743 .007
INSTTUIT -.863 .501 -.188 -1.724 .086
Mississippi, 4-Year Public
Entry model: ENROLL= HEAPP + HSGRAD + INSTTUIT + T + TT + TTT+ UNEMP
Final model: ENROLL = HEAPP + INSTTUIT + T + TT + CONSTANT
The significance of F (4,111) = 27.045, p<.001 is above the a< .05 threshold.
in enrollment is explained by in-state tuition (INSTTUIT), the short term (TT) impact of
the policy adoption, higher education appropriations (HEAPP) and time (T). Almost 52
125
coefficients (beta) estimates relative predictive power of in-state tuition (INSTTUIT), the
short term (TT) impact of the policy adoption, higher education appropriations (HEAPP)
and time (T). The impact of time (T) on enrollment was negative and significant. The
impact of average in-state tuition (INSTTUIT) and higher education appropriations
(HEAPP) were positive and significant on enrollment in 4-year public institutions in
Mississippi. The short (TT) and long (TTT) term impact of the policy adoption was
positive but not significant at the a< .05 level so accept He.
Table 43
Regression Analysis Coefficients, Final Model, for Mississippi, 4-Year Public3(N=l 12)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant) -2653.152 433.144 -6.125 .000
HEAPP 1.92E-006 .000 .497 3.323 .001
T -259.571 32.883 -1.702 -7.894 .000
TT 329.348 176.127 .257 1.870 .064
INSTTUIT 1.484 .143 1.337 10.386 .000
126
Mississippi, 2-Year Private
Final model: ENROLL = INSTTUIT + UNEMP + CONSTANT
Table 44
Regression Analysis Coefficients, Final Model, for Mississippi, 2-Year Private3 (N=23)
(Constant) 222.349 119.723 1.857 .078
UNEMP 22.415 10.623 .361 2.110 .048
INSTTUIT -.049 .016 -.513 -2.997 .007
enrollment is explained by in-state tuition (INSTTUIT) and unemployment (UNEMP).
Almost 43 percent of the variance is due to other factors other than the variables included
127
unemployment (UNEMP). The impact of average in-state tuition (INSTTUIT) was
negative and significant on enrollment in 2-year private institutions in Mississippi.
Unemployment (UNEMP) had a significant positive impact on enrollment. Accept the
H0 since there was no significant long (TTT) or short (TT) term impact on enrollment.
Mississippi, 4-Year Private
The significance of F (2,137) = 61.203, p<.001 is above the a< .05 threshold.
in enrollment is explained by time (T) and the average in-state tuition (INSTTUIT).
Almost 53 percent of the variance is due to other factors other than the variables included
coefficients (beta) estimates relative predictive power of time (T) and the average in-state
tuition (INSTTUIT). The impact of time (T) on enrollment was negative and significant.
The impact of average in-state tuition (INSTTUIT) was positive and significant on
enrollment in 4-year private institutions in Mississippi. Accept H0 since there was no
significant long (TTT) or short (TT) term impact.
128
Table 45
Regression Analysis Coefficients, Final Model, for Mississippi, 4-Year Private3(N=138)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant) 38.482 18.938 2.032 .044
T -6.420 2.081 -.205 -3.086 .002
INSTTUIT .026 .002 .734 11.032 .000
New Mexico, 2-Year Public
129
time (T). Accept H0 since there was no significant long (TTT) or short (TT) term impact
o f the policy adoption on enrollment.
Table 46
Regression Analysis Coefficients, Final Model, for New Mexico, 2-Year Public3(N=30)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant) -5401.529 1173.729 -4.602 .000
INSTTUIT 8.064 1.318 1.192 6.120 .000
T -423.139 92.910 -.887 -4.554 .000
3Dependent Variable: First time undergraduate enrollment (ENROLL), a ==.05
New Mexico. 4-Year Public
Ha: bb - ba 4- 0, the policy adoption did have a significant effect on enrollment.
Final model: ENROLL = INSTTUIT + HSGRAD + CONSTANT
130
Table 47
Regression Analysis Coefficients, Final Model, for New Mexico, 4-Year Public* (N=30)
(Constant)
-445.543 412.129
-1.081 .289
HSGRAD -.095 .034 -.249 -2.754 .010
INSTTUIT 1.660 .147 1.019 11.271 .000
in enrollment is explained by in-state tuition (INSTTUIT) and the number of high school
graduates (HSGRAD). Almost 17 percent of the variance is due to other factors other
in this state. The standard coefficients (beta) estimates relative predictive power of in
state tuition (INSTTUIT) and the number of high school graduates (HSGRAD). In-state
tuition (INSTTUIT) had a significant and positive impact on enrollment in 4-year
institutions, whereas the number of high school graduates (HSGRAD) had a negative and
significant impact on 4-year public enrollment. Accept H0 since there was no significant
long (TTT) or short (TT) term impact of the policy adoption on enrollment.
131
New Mexico. 2-Year Private
Ha: bb - ba # 0, the policy adoption did have a significant effect on enrollment.
Final model: ENROLL = TT + UNEMP + CONSTANT
Table 48
Regression Analysis Coefficients, Final Model, for New Mexico, 2-Year Private (N=9)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant) 1592.601 325.982 4.886 .003
TT -401.950 87.866 -1.335 -4.575 .004
UNEMP -180.014 45.040 -1.166 -3.997 .007
standard error for the final model. Nearly 71 percent (R2 =.710) of the variation in
enrollment is explained the immediate impact (TT) of the policy adoption and
132
unemployment (UNEMP). Almost 29 percent of the variance is due to other factors other
in this state. The standard coefficients (beta) estimates relative predictive power of the
immediate impact of the policy adoption (TT) and unemployment (UNEMP). Both the
immediate impact (TT) of the policy adoption and unemployment (UNEMP) had a
significant negative impact on enrollment in 2-year private institutions. Reject H0 since
there was a significant short (TT) term impact of the policy adoption on enrollment.
New Mexico 4-Year Private
The New Mexico 4-year private institution models did not yield any significant
results because there were not enough institutions with reported data in this category in
the state to analyze.
Nevada. 2-Year Public
Nevada, 2-year public did not yield a significant model for the effect of the
adoption of the merit aid policy on enrollment.
Nevada, 4-Year Public
The significance of F (1, 364) = 4.747, p=.030 is below the a< .05 threshold. The
133
in enrollment is explained by the average in-state tuition (INSTTUIT). Almost 99
coefficients (beta) estimates relative predictive power of the impact of the average in
state tuition (INSTTUIT), which had a significant and negative impact on enrollment.
Accept H0 since there was no significant long (TTT) or short (TT) term impact of the
Table 49
Regression Analysis Coefficients, Final Model, for Nevada, 4-Year Public8 (N=365)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant)
INSTTUIT
511.935 37.125
-.016 .007 -.114
13.789
-2.179
.000
.030
Nevada, 2-Year Private
There are no 2-year private institutions in Nevada.
Nevada, 4-Year Private
134
Table 50
Regression Analysis Coefficients, Final Model, for Nevada, 4-Year Private3(N=42)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant) 43.292 31.585 1.371 .178
INSTTUIT .006 .002 .410 2.841 .007
The significance of F (1, 41) = 8.070, p=.007 is below the a< .05 threshold. The
in enrollment is explained by in-state tuition (INSTTUIT). Almost 85 percent of the
estimates relative predictive power of in-state tuition (INSTTUIT). Accept H0 since there
was no significant long (TTT) or short (TT) term impact of the policy adoption on
enrollment.
135
South Carolina, 2-Year Public
Final model: ENROLL = INSTTUIT + T + UNEMP + CONSTANT
Table 51
Regression Analysis Coefficients, Final Model, for South Carolina, 2-Year Public3
(N=252)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant) 248.777 181.854 1.368 .173
INSTTUIT -.299 .088 -.386 -3.390 .001
T 71.968 18.067 .423 3.983 .000
UNEMP 89.310 35.897 .180 2.488 .014
Note: R= .251, R Square =.063, Adjusted R Square = .052.
error for the final model. Nearly 7 percent (R =.065) of the variation in enrollment is
136
explained by the average in-state tuition (INSTTUIT), time (T), and unemployment
(UNEMP). Almost 93 percent of the variance is due to other factors other than the
state. The standard coefficients (beta) estimates relative predictive power of the impact
of the average in-state tuition (INSTTUIT), time (T), and unemployment (UNEMP).
Both time (T) and unemployment (UNEMP) had a significant positive effect on
enrollment, while the average in-state tuition (INSTTUIT) had a significant negative
impact. Accept H0 since there was no significant long (TTT) or short (TT) term impact
of the policy adoption on enrollment.
South Carolina. 4-Year Public
Analysis of the 4-year public sector in South Carolina did not yield a significant
model.
South Carolina, 2-Year Private
Final model: ENROLL - INSTTUIT + T + CONSTANT
standard error for the final model. Almost 86 percent (adjusted R2 =.860) of the variation
in enrollment is explained by the average in-state tuition (INSTTUIT) and time (T).
Nearly 14 percent of the variance is due to other factors other than the variables included
137
state tuition (INSTTUIT) and time (T). The time (T) trend for enrollment in this sector
was significant and negative. Average in-state tuition (INSTTUIT) had a significant
positive effect on enrollment in the 2-year private sector. Accept H0 since there was no
Table 52
Regression Analysis Coefficients, Final Model, for South Carolina, 2-Year Private3
(N=17)
Unstandardized Standardized
t Sig.
(Constant) 43.906 38.511 1.140 .273
INSTTUIT .057 .006 .968 10.005 .000
T -12.912 3.840 -.325 -3.362 .005
Note: R= .937, R Square =.878, Adjusted R Square = .860.
South Carolina. 4-Year Private
138
Entry model: ENROLL^ T + TT + TTT+ HEAPP + HSGRAD + UNEMP + INSTTUIT
Final model: ENROLL= INSTTUIT + CONSTANT
Table 53
Regression Analysis Coefficients, Final Model, for South Carolina, 4-Year Private3
(N=269)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant)
INSTTUIT
227.690 30.976
.006 .003 .141
7.350
2.328
.000
.021
Note: R= .141, R Square =.020 , Adjusted R Square = .016
The significance of F (1, 268) = 5.419, p<.021, which is below the a< .05
threshold. Table 53 provides the unstandardized coefficients, standardized coefficients,
and standard error for the final model. Nearly 2 percent (adjusted R2=.016) of the
variation in enrollment is explained by the average in-state tuition (INSTTUIT). Almost
state tuition (INSTTUIT), which had a significant and positive impact on enrollment.
139
Accept H0 since there was no significant long (TTT) or short (TT) term impact o f the
Tennessee. 2-Year Public
All of the Tennessee analyses were done without the long term (TTT) dummy
variable because the policy has only been in place for one year.
Entry model: ENROLL= T + TT + HEAPP + HSGRAD + UNEMP + INSTTUIT
Table 54
Regression Analysis Coefficients, Final Model, for Tennessee, 2-Year Public3 (N=256)
Unstandardized
Coefficients
Standardized
Coefficients
t
Sig.
(Constant)
INSTTUIT
368.685 189.886
.384 .123 .316
1.942
3.130
.055
.002
Note: R= .316, R Square =100 , Adjusted R Square = .090
The significance of F (1,89) = 9.794, p<.002, which is below the a< .05
140
>y
and standard error for the final model. Nearly 9 percent (adjusted R =.090) of the
state tuition (INSTTUIT). Accept H0 since there was not a significant long (TTT) or
short (TT) term impact of the policy adoption on enrollment.
Tennessee. 4-Year Public
The analysis of Tennessees 4-year public sector did not yield a significant model.
Tennessee. 2-Year Private
Final model: ENROLL= INSTTUIT + T + CONSTANT
coefficients, and standard error for the final model. Nearly 29 percent (adjusted R2
=.293) of the variation in enrollment is explained by the average in-state tuition
(INSTTUIT) and time (T). Almost 71 percent of the variance is due to other factors other
141
in this state. The standard coefficients (beta) estimates relative predictive power of the
impact of the average in-state tuition (INSTTUIT) and time (T). Accept H0 since there
was not a significant long (TTT) or short (TT) term impact of the policy adoption on
enrollment.
Table 55
Regression Analysis Coefficients, Final Model, for Tennessee, 2-Year Private3(N=19)
(Constant) 102.725 40.777 2.519 .023
T -15.864 7.462 -.432 -2.126 .049
INSTTUIT .017 .006 .534 2.629 .018
Tennessee. 4-Year Private
142
Table 56
Regression Analysis Coefficients, Final Model, for Tennessee, 4-Year Private3 (N=255)
(Constant) -32.871 39.052 -.842 .401
T -18.235 7.130 -.132 -2.558 .011
INSTTUIT .035 .003 .630 12.225 .000
and standard error for the final model. Nearly 37 percent (adjusted R2 =.368) of the
variation in enrollment is explained by the average in-state tuition (INSTTUIT) and time
(T). Almost 63 percent of the variance is due to other factors other than the variables
included in the entry model which were stepped out as not significant in this state. The
standard coefficients (beta) estimates relative predictive power of the impact of the
average in-state tuition (INSTTUIT) and time (T). Accept Hc since there was not a
143
Washington. 2-Year Public
Final model: ENROLL= HSGRAD + HEAPP + T + TT + TTT + UNEMP +
CONSTANT
coefficients, and standard error for the final model. Nearly 55 percent (adjusted R2
=.546) of the variation in enrollment is explained by higher education appropriations
(HEAPP), the number of high school graduates (HSGRAD), unemployment (UNEMP),
the long (TTT) and short (TT) term impact of the policy adoption, and time (T). Almost
coefficients (beta) estimates relative predictive power of the impact of time (T), which
had a significant and negative effect on enrollment; the higher education appropriations
(HEAPP), the number of high school graduates (HSGRAD), the long (TTT) and short
(TT) term impact of the policy adoption, and unemployment (UNEMP), all of which had
significant and positive impact on enrollment. Reject H0 since there was a significant
long (TTT) and short (TT) term impact of the policy adoption on enrollment.
144
Table 57
Regression Analysis Coefficients, Final Model, for Washington, 2-Year Public3(N=256)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant) -46176.45 5568.368 -8.293 .000
HEAPP 2.65E-005 .000 3.212 8.533 .000
HSGRAD .210 .040 .700 5.205 .000
T -3285.565 313.007 -8.380 -10.497 .000
TT 897.969 425.776 .431 2.109 .036
TTT 3152.241 312.608 2.455 10.084 .000
UNEMP 2424.117 215.873 2.786 11.229 .000
Washington, 4-Year Public
145
Table 58
Regression Analysis Coefficients, Final Model, for Washington, 4-Year Publica (N=48)
(Constant) -5390.703 783.430 -6.881 .000
INSTTUIT 2.825 .303 1.258 9.339 .000
T -489.572 77.251 -.854 -6.337 .000
The significance of F (2, 47) = 44.439, pc.001, which is below the a< .05
coefficients, and standard error for the final model. Nearly 65 percent (adjusted R2 =.649)
of the variation in enrollment is explained by the average in-state tuition (INSTTUIT)
and time (T). Almost 35 percent of the variance is due to other factors other than the
state. The standard coefficients (beta) estimates relative predictive power of the impact
of average in-state tuition (INSTTUIT) and time (T). The average in-state tuition
(INSTTUIT) had a significant and positive effect on enrollment. Time (T) had a
significant negative impact on enrollment. Accept H0 since there was no significant long
(TTT) or short (TT) term impact of the policy adoption on enrollment.
146
Washington, 2 Year Private
Washington, 2 year private had N=2, which was not enough observations to
analyze the impact of the policy adoption in that sector.
Washington, 4-Year Private
Final model: ENROLL= INSTTUIT + HEAPP + CONSTANT
coefficients, and standard error for the final model. Almost 51 percent (adjusted R2
=.513) of the variation in enrollment is explained by higher education appropriations
(HEAPP) and the average in-state tuition (INSTTUIT). Nearly 49 percent of the variance
is due to other factors other than the variables included in the entry model which were
stepped out as not significant in this state. The standard coefficients (beta) estimates
relative predictive power of the impact of average in-state tuition (INSTTUIT) and the
higher education appropriations (HEAPP). Higher education appropriations (HEAPP)
had a negative and significant effect on enrollment in the 4-year private institutions. The
average in-state tuition (INSTTUIT) had a significant and positive effect on enrollment.
Accept H0 since there was no significant long (TTT) or short (TT) term impact of the
147
Table 59
Regression Analysis Coefficients, Final Model, for Washington, 4-Year Private3(N=120)
(Constant) 459.249 194.783 2.358 .020
HEAPP -5.53E-007 .000 -.236 -3.501 .001
INSTTUIT .037 .003 .761 11.290 .000
West Virginia. 2-Year Public
Final model: ENROLL= INSTTUIT + TTT + HEAPP + CONSTANT
coefficients, and standard error for the final model. Nearly 29 percent (adjusted R2=.294)
of the variation in enrollment is explained by the average in-state tuition (INSTTUIT),
the long term (TTT) impact of the policy adoption, and higher education appropriations
(HEAPP). Almost 71 percent of the variance is due to other factors other than the
148
state. The standard coefficients (beta) estimates relative predictive power of the average
in-state tuition (INSTTUIT), the long term (TTT) impact of the policy adoption, and
higher education appropriations (HEAPP). The average in-state tuition (INSTTUIT) had
a significant negative effect on enrollment. The long term (TTT) impact of the policy
adoption and higher education appropriations (HEAPP) both had a negative and
significant impact on 2-year public enrollment in West Virginia. Reject H0 since there
was a significant long (TTT) term impact of the policy adoption on enrollment.
Table 60
Regression Analysis Coefficients, Final Model, for West Virginia, 2-Year Public3(N=43)
(Constant) 3263.136 935.269 3.489 .001
HEAPP -6.95E-006 .000 -.496 -2.840 .007
INSTTUIT -.193 .064 -.408 -3.030 .004
TTT -118.554 63.658 -.325 -1.862 .070
West Virginia, 4-Year Public
149
Table 61
Regression Analysis Coefficients, Final Model, for West Virginia, 4-Year Public3(N=88)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant)
INSTTUIT
261.441 209.401
.352 .100 .356
1.249
3.532
.215
.001
coefficients (beta) estimates relative predictive power of the impact of average in-state
tuition (INSTTUIT). The average in-state tuition had a significant and positive effect on
150
enrollment. Accept H0 since there was no significant long (TTT) or short (TT) term
impact o f the policy adoption on enrollment.
West Virginia. 2-Year Private
No significant model could be developed because there were only two
observations for West Virginias 2-year private institutions.
West Virginia. 4-Year Private
West Virginia, 4-year private did not yield any significant models from the
analysis.
Research Question 4 (RQ4)
Research Question 4 (RQ4): What is the effect of the following three types of
merit aid programs across similar programs by sector and level: a. Full tuition; b. Partial
tuition; c. One time payment?
Full Tuition Program. 2-Year Public
Ha: bb - ba f 0, the policy adoption did have a significant effect on enrollment.
Final model: ENROLL= INSTTUIT + HSGRAD + HEAPP + UNEMP + CONSTANT
and standard error for the final model. Almost 22 percent (adjusted R2=.223) of the
variation in enrollment is explained by the average in-state tuition (INSTTUIT), the
151
number of high school graduates (HSGRAD), higher education appropriations (HEAPP),
and unemployment (UNEMP). Nearly 78 percent of the variance is due to other factors
power of the average in-state tuition (INSTTUIT), higher education appropriations
(HEAPP), the number of high school graduates (HSGRAD), and unemployment
(UNEMP). The long (TTT) and short (TT) term impact of the policy adoption were both
stepped out of the equation and had no significant impact on enrollment across 2-year
public institutions in states with full merit scholarship programs therefore accept H0.
Table 62
Regression Analysis Coefficients, Final Model, for 2-Year Public Institutions Across Full
Tuition Payment Merit Scholarship States8 (N=1960)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant) -992.339 150.184 -6.607 .000
HEAPP 2.70E-007 .000 .201 2.992 .003
HSGRAD .009 .002 .275 4.476 .000
INSTTUIT .164 .040 .089 4.124 .000
UNEMP 120.711 23.301 .129 5.181 .000
152
Full Tuition Program, 4-year public
Final model: ENROLL= INSTTUIT + TTT + TT + T + HEAPP + CONSTANT
the long (TTT) and short (TT) term impact of the policy adoption, time (T), and higher
education appropriations (HEAPP). Almost 69 percent of the variance is due to other
power of the average in-state tuition (INSTTUIT), the long (TTT) and short (TT) term
impact of the policy adoptions, time (T), and unemployment (UNEMP). The long term
(TTT) impact of the policy adoption had a positive but not significant impact on
enrollment across 4-year public institutions in states with full merit scholarship programs.
Higher education appropriations (HEAPP), the average in-state tuition (INSTTUIT) and
the short term (TT) impact of the policy adoption all had a significant positive effect on
enrollment across 4-year public institutions in states that had full tuition payment merit
aid programs. Time (T) had a significant and negative impact on enrollment in this
153
sector. Reject H0 since there was a significant short (TT) term impact of the policy
Table 63
Regression Analysis Coefficients, Final Model, for 4-Year Public Institutions Across Full
Tuition Payment Merit Scholarship States3(N=765)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant) -460.720 194.186 -2.373 .018
HEAPP 1.06E-006 .000 .548 15.001 .000
INSTTUIT .802 .081 .323 9.842 .000
T -143.255 23.829 -.519 -6.012 .000
TT 259.746 130.915 .098 1.984 .048
TTT 59.556 33.724 .147 1.766 .078
Full Tuition Program. 2-Year Private
154
Final model: ENROLL= HSGRAD + T + TTT + CONSTANT
Table 64
Regression Analysis Coefficients, Final Model, for 2-Year Private Institutions Across
Full Tuition Payment Merit Scholarship States3(N=208)
(Constant) 83.482 53.514 1.560 .120
HSGRAD .001 .001 .128 1.802 .073
T 13.148 6.244 .411 2.106 .036
TTT -17.812 8.372 -.414 -2.128 .035
The significance of F (4, 207) = 2.453, p=.047, which is below the a< .05
of the variation in enrollment is explained by the number of high school graduates
(HSGRAD), time (T), and the long term (TTT) impact of the merit programs. Almost 98
coefficients (beta) estimates relative predictive power of the number of high school
155
graduates (HSGRAD), the long term (TTT) impact of the policy adoptions, and time (T).
The long term (TTT) impact of the policy adoption had a significant negative impact on
enrollment across 2-year private institutions in states with full merit scholarship
programs. The number of high school graduates (HSGRAD) had a positive, but not
significant, effect on enrollment. Time (T) had a significant positive effect on
enrollment. Reject H0 since there was a significant long (TTT) term impact of the policy
Full Tuition Program. 4-Year Private
Final model: ENROLL= INSTTUIT + HEAPP + HSGRAD + TT + UNEMP +
CONSTANT
and standard error for the final model. Nearly 24 percent (adjusted R =.235) of the
number of high school graduates (HSGRAD), higher education appropriations (HEAPP),
unemployment (UNEMP), and the short term (TT) impact of the merit programs. Almost
entry model which were stepped out as not significant in this state.
156
Table 65
Regression Analysis Coefficients, Final Model, for 4-year Private Institutions Across Full
Tuition Payment Merit Scholarship States8 (N=l 175)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant) 168.017 78.743 2.134 .033
HEAPP -3.09E-007 .000 -.706 -5.725 .000
HSGRAD .007 .001 .619 5.695 .000
INSTTUIT .033 .002 .503 18.502 .000
TT 63.875 33.393 .087 1.913 .056
UNEMP -37.243 12.337 -.109 -3.019 .003
The standard coefficients (beta) estimates relative predictive power of the average
in-state tuition (INSTTUIT), the number of high school graduates (HSGRAD), the short
term impact (TT) of the policy adoptions, and unemployment (UNEMP). The short term
(TT) impact of the policy adoption had a positive but not significant impact on
enrollment across 4-year private institutions in states with full merit scholarship
programs. Higher education appropriations (HEAPP) and the unemployment rate
(UNEMP) both had a significant negative impact on enrollment. Finally, the number of
high school graduates (HSGRAD) and the average in-state tuition (INSTTUIT) both had
157
a significant positive effect on enrollment across 4-year private institutions in states that
had full tuition payment merit aid programs. Accept H0 since there was no significant
long (TTT) or short (TT) term impact of the policy adoption on enrollment.
Partial Tuition Program. 2-Year Public
Final model: ENROLL= HEAPP + T + TT + TTT + HSGRAD+ CONSTANT
of the variation in enrollment is explained by the higher education appropriations
(HEAPP), the number of high school graduates (HSGRAD), time (T), and the short (TT)
and long (TTT) term impact of the policy adoption. Almost 96 percent of the variance is
due to other factors other than the variables included in the entry model which were
relative predictive power of the higher education appropriations (HEAPP), the number of
high school graduates (HSGRAD), time (T), and the short (TT) and long (TTT) term
impact of the policy adoption. Higher education appropriations (HEAPP) and the short
(TT) and long (TTT) term impact of the policy adoption all had a significant positive
impact on enrollment in 2-year public institutions in states with partial tuition payment
merit aid programs. The number of high school graduates (HSGRAD) and time (T) both
158
had a significant negative effect on enrollment. Reject H0 since there was a significant
long (TTT) and short (TT) term impact of the policy adoption on enrollment.
Table 66
Regression Analysis Coefficients, Final Model, for 2-Year Public Institutions Across
Partial Tuition Payment Merit Scholarship States3(N=1576)
(Constant) 1151.449 74.737 15.407 .000
HEAPP 5.02E-007 .000 .206 5.017 .000
HSGRAD -.013 .002 -.240 -6.051 .000
T -105.694 17.027 -.594 -6.207 .000
TT 155.230 59.514 .110 2.608 .009
TTT 105.823 18.666 .472 5.669 .000
Partial Tuition Program. 4-Year Public
159
Final model: ENROLL= HEAPP + INSTTUIT + T + CONSTANT
Table 67
Regression Analysis Coefficients, Final Model, for 4-Year Public Institutions Across
Partial Tuition Payment Merit Scholarship States3(N=854)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant) 553.409 97.376 5.683 .000
HEAPP 6.95E-007 .000 .214 5.974 .000
INSTTUIT .156 .031 .190 4.972 .000
T -27.489 7.417 -.129 -3.706 .000
time (T), and higher education appropriations (HEAPP). Almost 89 percent of the
estimates relative predictive power of the average in-state tuition (INSTTUIT), time (T),
160
and higher education appropriations (HEAPP). Higher education appropriations
(HEAPP) and the average in-state tuition (INSTTUIT) both had a significant positive
effect on enrollment across 4-year public institutions in states that had partial tuition
payment merit aid programs. Time (T) had a significant negative impact on enrollment
in this sector. Accept H0 since there was no significant long (TTT) or short (TT) term
Partial Tuition Program. 2-Year Private
Final model: ENROLL= INSTTUIT + TT + CONSTANT
of the variation in enrollment is explained by the average in-state tuition (INSTTUIT)
and the short (TT) term impact of the policy adoption. Almost 87 percent of the variance
is due to other factors other than the variables included in the entry model which were
relative predictive power of the average in-state tuition (INSTTUIT) and the short term
(TT) impact of the policy adoption. Average in-state tuition (INSTTUIT) had a
significant positive impact on enrollment in 2-year private institutions in states with
partial tuition payment merit aid programs. The short term (TT) impact of the merit aid
161
policies was significantly negative. Reject H0 since there was a significant short (TT)
term impact o f the policy adoption on enrollment.
Table 68
Linear Regression Analysis Coefficients, Final Model, for 2-Year Private Institutions
Across Partial Tuition Payment Merit Scholarship States3(N=197)
Unstandardized Standardized t
Sig.
(Constant) 87.023 21.914 3.971 .000
INSTTUIT .018 .003 .365 5.425 .000
TT -43.555 21.809 -.134 -1.997 .047
Partial Tuition Program. 4-year private
Final model: ENROLL= UNEMP + HEAPP + INSTTUIT + HSGRAD + CONSTANT
162
Table 69
Regression Analysis Coefficients, Final Model, for 4-Year Private Institutions Across
Partial Tuition Payment Merit Scholarship States8 (N=1687)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant) 189.446 42.127 4.497 .000
HEAPP -2.02E-007 .000 -.207 -6.862 .000
HSGRAD .002 .001 .096 3.408 .001
INSTTUIT .024 .001 .483 20.099 .000
UNEMP -13.062 5.369 -.057 -2.433 .015
number of high school graduates (HSGRAD), unemployment (UNEMP), and higher
education appropriations (HEAPP). Almost 79 percent of the variance is due to other
power of the average in-state tuition (INSTTUIT), the number of high school graduates
163
(HSGRAD), unemployment (UNEMP), and higher education appropriations (HEAPP).
Higher education appropriations (HEAPP) and the unemployment rate (UNEMP) both
had a significant negative effect on enrollment across 4-year private institutions in states
that had partial tuition payment merit aid programs. Average in-state tuition
(INSTTUIT) and the number of high school graduates (HSGRAD) both had a significant
positive effect on enrollment across 4-year private institutions in the partial payment
merit aid states. Accept H0 since there was no significant long (TTT) or short (TT) term
One Time Payment. 2-Year Public
standard error for the final model. More than 17 percent (adjusted R2 =.171) of the
variation in enrollment is explained by average in-state tuition (INSTTUIT),
unemployment (UNEMP), and the number of high school graduates (HSGRAD). Almost
(INSTTUIT), unemployment (UNEMP), and the number of high school graduates
164
(HSGRAD). The impact of average in-state tuition (INSTTUIT) was negative and
significant. The impact of the number of high school graduates (HSGRAD) and
unemployment (UNEMP) were both positive and significant for 2-year public
institutions. Accept H0 since there was no significant long (TTT) or short (TT) term
Table 70
Regression Analysis Coefficients, Final Model, for One Time Payment, 2-Year Public3
(N=249)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant) -1864.295 400.342 -4.657 .000
HSGRAD .051 .008 .450 6.740 .000
UNEMP 66.326 25.843 .151 2.566 .011
INSTTUIT -.188 .033 -.376 -5.660 .000
One Time Payment. 4-vear public
Ha: bb - ba i 10, the policy adoption did have a significant effect on enrollment.
165
Table 71
Regression Analysis Coefficients, Final Model, for One Time Payment, 4-Year Public3
(N=168)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant) 581.529 228.863 2.541 .012
T -46.305 24.739 -.176 -1.872 .063
INSTTUIT .326 .085 .361 3.829 .000
error for the final model. More than 7 percent (adjusted R2 =.072) of the variation in
enrollment is explained by average in-state tuition (INSTTUIT) and time (T). Nearly 93
(INSTTUIT) and time (T). The impact of the time (T) on enrollment was negative but
166
not significant and the impact of average in-state tuition (INSTTUIT) was positive and
significant for 4-year institutions. Accept H0 since there was no significant short (TT) or
long (TTT) term impact on enrollment.
One Time Payment, 2-Year Private
Final model: ENROLL = HSGRAD + T + CONSTANT
in enrollment is explained by time (T) and the number of high school graduates
(HSGRAD). Almost 82 percent of the variance is due to other factors other than the
state. The standard coefficients (beta) estimates relative predictive power of time (T) and
the number of high school graduates (HSGRAD). The impact of high school graduates
(HSGRAD) on enrollment was significantly negative while the impact of time was
positive and significant for 2-year private institutions. Accept H0 since there was no
167
Table 72
Regression Analysis Coefficients, Final Model, for One Time Payment, 2-year private3
(N=65)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant) 3066.463 1344.091 2.281 .026
HSGRAD -.066 .030 -1.033 -2.203 .031
T 73.684 29.051 1.189 2.536 .014
One Time Payment. 4-Year Private
Final model: ENROLL = HEAPP + INSTTUIT + UNEMP + CONSTANT
standard error for the final model. More than 27 percent (adjusted R2 =.272) of the
variation in enrollment is explained by average in-state tuition (INSTTUIT),
unemployment (UNEMP), and higher education appropriations (HEAPP). Nearly 72
168
coefficients (beta) estimates relative predictive power of unemployment (UNEMP),
average in-state tuition (INSTTUIT), and higher education appropriations (HEAPP). The
impact of the higher education appropriations (HEAPP) and unemployment (UNEMP) on
four year private enrollment were significant and negative. However, the impact of the
average in-state tuition (INSTTUIT) was positive and significant. Accept H0 since there
was no significant long (TTT) or short (TT) term impact on enrollment.
Table 73
Regression Analysis Coefficients, Final Model, for One time payment, 4-year private3
(N=444)
Unstandardized
Coefficients
Standardized
Coefficients
t Sig.
(Constant) 561.586 169.366 3.316 .001
HEAPP -4.44E-007 .000 -.231 -3.876 .000
UNEMP -55.877 20.026 -.162 -2.790 .005
INSTTUIT .037 .003 .555 12.987 .000
169
Summary
Chapter 5 presented the graphs of the time series data and the results of the time
series regression analyses of the data. In all, the longitudinal data for each sector in the
15 states in this study were graphed to determine the onset and duration of the impact the
policy adoption had. The analysis consisted of a backward multiple linear regressions to
identify the predictive value and relationship between the merit aid policy adoption and
first time undergraduate enrollment in the 15 states that have adopted these policies and
across the three types of merit aid programs.
Based on Longs (2003) work on the impact of the Georgia HOPE Scholarship on
institutions and tuition, three variables were identified as having significant impact: state
higher education appropriations, unemployment, and average tuition for in-state students.
Because the number of high school graduates in a state is directly linked to the number of
potential candidates for undergraduates, this study also accounts for the number of public
high school graduates in each state.
Research question 3 was an examination of the impact of the policy adoption on
the four sectors within each state that has a merit aid policy. The policy adoption had a
significant short term impact on enrollment in four cases. In Alaska and Washington,
there was a significant positive relationship between the adoption of the policy at the state
level and short term enrollment in 2-year public institutions. In Kentucky, there was a
significant positive relationship between the policy adoption and short term enrollment in
2-year private institutions in that state. And, in Nevada there was a negative relationship
between the policy adoption and the short term enrollment at the 2-year private sector
institution.
170
There was a significant long term impact of the policy adoption on 2-year public
institutions in four cases. In Alaska, Louisiana, and West Virginia, there was a
significant negative relationship between the policy adoption and enrollment in the 2-year
public institutions. In Washington, there was a significant positive relationship between
the policy adoption and enrollment in the 2-year public intuitions. There was a
significant long term impact on 2-year private institutions in two cases. In Arkansas,
there was a negative relationship between the policy adoption and enrollment at 2-year
private institutions. In Florida, there was a positive relationship between the
implementation of the policy and enrollment at 2-year privates. The impact on 4-year
public institutions interestingly only had a significant impact in one state, Missouri,
which had a negative relationship. The long term impact of the policy adoption on 4-year
private institutions was significant in only three states, Arkansas, Florida, and West
Virginia, all three of which were negative.
Research question 4 was an examination of the impact of merit aid programs
across similar programs by sector and level. There was no significant long or short term
impact on enrollment in 2-year public or 4-year private institutions after the policy
adoption of full payment tuition plans. Full tuition merit scholarship programs did have a
significant long term negative effect on enrollment at 2-year private institutions across
the six states with these programs. 4-year public institutions were significantly and
positively impacted by the full tuition merit aid policies.
Partial tuition programs did not have a significant impact on enrollment in 4-year
public and private institutions. There was no significant long or short term impact on
enrollment in 2-year private institutions after the adoption of partial tuition payment
171
plans. There was a significant and positive long and short term effect at 2-year public
institutions across the six states that have adopted the partial payment programs.
Finally, in the state that had a one time payment program, there was no significant
effect on enrollment in any of the sectors (2, 4-year, public or private).
172
CHAPTER 6
SUMMARY, CONCLUSIONS, AND RECOMMENDATIONS
This final chapter begins with a broad presentation of each research question and
broad interpretation of the results. Then, the findings are supported using specific state
examples or results. Finally, implications for higher education policy are discussed as are
recommendations for future study. For this study, Tong term effect means that there was
a significant impact on enrollment from year-to-year after the policy adoption, whereas
the immediate impact or short term effect is the impact of the policy adoption on just
the first year after implementation.
Summary of Results
Research question 1: How has merit aid impacted first time undergraduate
policy?
Research question one (RQ1) is a state level analysis of how merit aid has
impacted first time undergraduate postsecondary participation in each of the states that
have adopted these student financial aid policies. The overall impact of the adoption of
the merit aid policies is that they have had a significant effect on enrollment across nine
of the fifteen states in this study. Specifically, they have had a significant positive,
immediate effect on enrollment across Florida, Tennessee, and West Virginia but a
173
significant negative, immediate effect on enrollment in Nevada and Alaska. The long
term effect has been significant and positive in Alaska, Florida, Kentucky, Louisiana,
Missouri, and New Mexico but significantly negative in Arkansas, (see Appendix D for
full results). There was no significant long or short term effect in only three states,
Georgia, Michigan, Mississippi, and South Carolina.
Merit aid programs, at a state level, seem to have a positive impact on enrollment
which could be because they encourage participation somewhere in the system. Another
possibility is that as states prepare for these initiatives, there is generally more public
discussion about college costs, participation, and preparation. As issues rise to the
forefront of public discussion, the level of the general knowledge around the issue
increases. In this case, as more publicity surrounds the adoption of the merit aid policies,
students and parents are more likely to have discussions about postsecondary education
and paying for college.
In addition to the impact of the policy adoption, unemployment had an impact on
participation across six of the states in this study. Unemployment had a significant
negative impact in Nevada and Alaska and a significant positive effect in Arkansas,
Georgia, Mississippi, and South Carolina. It could be that there was a significant
negative relationship between unemployment and enrollment in Nevada and Alaska
because in these states, higher education is not necessarily associated with employment
as strongly as it is in other states. Particularly in Nevada, with a large proportion of the
population employed in the construction and service industries, education might not offer
the same incentive as it does in other states.
In South Carolina, there was no significant change in enrollment upon the
174
adoption of the first merit aid policy, which may explain why there was another program
adopted less than five years later. The second set of merit aid policies that was adopted
in South Carolina had a significant positive impact on enrollment. Where most states
have adapted their merit aid policies to allow for differential awards by sector, South
Carolina has taken the approach of writing new legislation for new award programs.
There are currently four different merit scholarships available in South Carolina. The
interaction effects of these four scholarship programs make it difficult to determine the
impact of any one of the programs at this broad level of analysis. One assumption of
time series analysis is that there are no other alternative explanations for the
phenomenon. In the case of South Carolina, there are clearly several different legislative
initiatives in higher education occurring simultaneously.
In Washington, there was not a significant effect on enrollment upon adoption of
the merit aid policy which may have been a factor in why the program ended on June 30,
2006. The award amount in Washington was $1,254.00, whereas the 2005 average 4-
year public tuition was $4,630; 2-year public was $2, 230; and 4-year private was
$18,300. This payment was significantly less than 50 per cent in terms of proportion of
tuition paid by the award, which may explain why students did not respond by enrolling
at significantly higher numbers.
There was one state where the adoption of the merit aid policy had a significant
negative long-term impact on enrollmentArkansas. It is evident from the graph of
enrollment in Arkansas that there are large fluctuations in enrollment from year to year.
There were five data points before the policy adoption and thirteen after, which may not
be sufficient data for enrollment trends pre-policy adoption. A longer pre-policy
175
adoption time series might show that over a longer period, the impact of the policy is in
fact positive or neutral. Arkansas requires that students take a core set o f courses in high
school, achieve a minimum grade point average, and ACT or SAT score. Additionally,
unlike most of the merit aid programs in other states, Arkansas has a $60,000 family
income cap for recipients with one child, and makes additional allowances for families
with more children (Arkansas Department of Higher Education, 2006). An unintended
consequence of the family income cap is that it could also contribute to the decline in
enrollment because students who are eligible in merit but exceed the income cap may be
selecting schools in other states because they did not receive the grant to stay in
Arkansas.
The short term impact of the merit aid policies was mixed; in Florida, Tennessee
and West Virginia, the adoption of the merit aid policies had a significant positive effect
on short-term enrollment. Tennessee and West Virginia adopted their policies less than
five years ago and so are only used as an indicator of the initial enrollment response. The
incentive effect of merit aid policies across the states appears to gain momentum rather
than to have a sudden consistent impact on college going. Part of the reason for the lag in
student enrollment response is that the criterion for an award is based on four years of
performance (grade point average) or a score on a test (ACT or SAT) taken in the junior
year, both events that took place before the legislation. In Alaska and Nevada, the short
term impact was significant and negative. As with all legislation, when merit aid policies
are adopted in states, financing does not necessarily immediately accompany the
legislation. In the case of merit aid, because these programs can be cumbersome and
expensive to implement, there can be a lag in the response by students because of
176
structural issues with the programs.
Research Question 2: What is the effect of the following three types of merit aid
programs on first time undergraduate enrollment across states (a) full tuition payment
programs, (b) partial tuition payment programs, (c) one time payment programs?
Full Tuition Payment Programs
In the aggregated states where there was a full tuition payment policy adopted,
there was no significant long or short term impact of the policy adoption on enrollment.
There is a significant increasing enrollment trend in the full tuition payment states over
time (T) that is unrelated to the policy adoption. Higher education appropriations
(HEAPP), the number of high school graduates (HSGRAD), and unemployment
(UNEMP) all significantly affected enrollment in the full tuition payment merit aid states.
Higher education appropriations are a reflection of the general fiscal welfare of the state
and its citizens. Earlier studies have shown that merit aid disproportionately benefits
students that come from advantaged backgrounds who are already likely to go to college
(Cornwell, Mustard, & Sridhar, 2003; Dynarski, 2003; Heller & Rasmussen, 2003). By
extension, the adoption of a merit aid policy might not have a significant enrollment
effect in states with higher appropriations. In states where full merit aid policies are
adopted, unemployment (UNEMP) had a significant positive effect on postsecondary
enrollment. In states where education is a clear vehicle to better paying jobs, going to
college right after high school makes sense. Since there is a significant positive
relationship between the number of high school graduates and undergraduate enrollment
in these states, it appears that students are doing just that.
177
Finally, in foil tuition payment policy states, the average in-state tuition
(INSTTUIT) had a significant negative effect on enrollment. There is a history of
research that supports the finding that as tuition rises, the number of students enrolling
decreases. What is interesting about this case is that the relationship persists even in
states where there is a full tuition payment program in place to defray tuition costs. In
her 2003 study of how Georgia institutions responded to the adoption of the HOPE
scholarship policy, Long found that institutions may have increased fees, room and
board, and other associated costs rather than just tuition. The linkage between rising
tuition and associated fees may be a phenomenon that is occurring across all of the full
tuition payment merit aid states.
Partial Tuition Payment Programs
The aggregated state level analysis found that the partial tuition payment
programs had no significant immediate impact but that they did have a significant
negative long term impact on enrollment. One possible explanation for the negative long
term results where the partial tuition programs are in place is that there is wide variation
between policies across the states as well as within the same state policy from year to
year. There are two main areas of variance across partial tuition payment merit aid
programsselection criteria and award amount. For example, in Nevada the criterion for
award was a 3.0 grade point average at a Nevada high school whereas in Washington,
students were required to be in the top 15% of their graduating class. That difference in
criteria alone would likely lead to significant differences in the number of students
eligible for the merit awards and the impact on enrollment, especially across states that
are pooled together for analytical purposes.
178
Another possible reason that the policy adoption across partial payment merit aid
states has had a negative impact on enrollment is that the eligibility requirements for the
merit aid scholarships have changed in almost every state since their adoption. One of
the reasons for the changes in eligibility criteria is that the money to pay for the
scholarships comes from limited resources such as tobacco dollars or lottery revenues,
with only a very few states using general revenues. As the number of students eligible
for the award increases faster than the revenues is growing, states are forced to make
decisions about how to stretch limited resources by either reducing the amount of award
or the number of eligible students.
Like the full tuition payment states, the partial tuition payment states also
demonstrated significant positive relationships between enrollment and higher education
appropriations, the number of high school graduates, and unemployment. It is likely that
the same reasons behind the positive relationship between enrollment and appropriations
(wealth), graduates (supply and preparation), and unemployment (incentives) in the full
tuition states are the same underlying causes of the significant relationships in the partial
tuition states.
One Time Payment
The analysis of the single one time payment state, Michigan, revealed that there
was not a significant long or short term impact of the policy adoption on enrollment. The
one time tuition payment in this state is only $2500, significantly less than the average
cost of tuition at a 4-year public institution in the state. Since the policy is a one time
payment, and it is the only state with that policy, it is not possible to determine i f that
alone is the reason that there was no incentive effect. However, in their 2003 study of
179
the incentive effect of Michigans merit aid policy, Heller and Rogers speculated that this
relatively low one time payment did not sufficiently off-set the cost of tuition enough to
serve as an incentive to enroll.
Research Question 3: How much has merit aid impacted participation by sector
and level in the states that have adopted this financial aid policy?
2-Year Public Sector
There was a significant increase in student enrollment in the 2-year public
institutions immediately after the adoption of the merit aid policies in Alaska and
Washington. Many of the states did not have any effect after the adoption of the merit
aid program. The 2-year public enrollment did not immediately significantly decline in
any of the states. However, there was a significant negative effect of the merit aid
adoption over a longer period in Alaska, Louisiana, and West Virginia. This is an
important finding because earlier work on merit aid suggested that it pulls students from
2-year institutions up to the 4-year institutions (Dynarski, 2002). This study suggests that
while that might be the case over long periods of time, it is not the immediate impact of
the program. Furthermore, i f students who were not inclined to enroll are more likely to
enroll at 2-year institutions because of merit aid, they could be taking the place of those
students who are now incentivized to move up to 4-year institutions.
Average in-state tuition (INSTTUIT) significantly positively influenced
enrollment in 2-year public institutions in Arkansas, Georgia, Kentucky, Louisiana, New
Mexico, and Tennessee. A likely reason is that as tuition in 4-year institutions raises,
students enroll in the less expensive 2-year institutions. Additionally, unemployment
(UNEMP) had a significant positive effect on enrollment in 2-year public institutions.
180
This finding is not surprising, as 2-year public institutions are often where workers turn
to retool for new jobs. Finally, since 2-year public institutions are a place where many
students begin their postsecondary educational journey, it is not surprising to find that the
number of high school graduates (HSGRAD) is also positively associated with 2-year
public enrollment in Georgia, Michigan, and Washington.
4-Year Public Sector
Across nearly every state, the adoption of the merit aid policies do not have a
significant short or long term impact on 4-year public enrollment. Part of the reason that
there was no significant effect of the merit aid adoption on enrollment in this sector may
be that many 4-year public institutions are at or near capacity and do not have room to
take significantly more students. Another reason may be that the students that benefit
from these programs are students that would enroll at 4-year public institutions regardless
of the merit aid adoption. In their examination of the New Mexico Lottery Scholarship
program, Binder, Ganderton, and Hutchens (2001) found that the merit aid policy did not
influence whether students went to college or not but rather where they went to college.
The average in-state tuition (INSTTUIT) has a significant positive impact on
enrollment in the 4-year sector in Arkansas, Georgia, Kentucky, Louisiana, Michigan,
Missouri, Mississippi, New Mexico, and West Virginia. As average in state tuition
increases in these nine states, the number of students that enroll in 4-year public
institutions also significantly increases. One reason could be that parents and students are
aware of the benefits of higher education and believe that the cost will not go down in the
future. Hoping to take advantage of current prices, students enroll despite rising tuition
costs. Another possibility simply concerns supply and demand. Student enrollment
181
demand for the 4-year sector in these states may be higher than the spaces available
(supply). An economic consequence of an imbalance whereby demand is greater than
supply is a rise in prices (tuition). If this is the case in these states, students demand is
inelastic relative to other states.
Nevada was the only state in which there was a significant negative effect of the
average in-state tuition on enrollment. Here again, Nevada is a state in which the
perceived return on higher education is less than that of other states so when tuition rises,
enrollment in the 4-year public institutions decreases.
The adoption of a merit aid policy had a significant negative long term effect on
enrollment in the 4-year public sector in Missouri. The merit scholarship program in
Missouri offers the top three percent of students who take the ACT or SAT in Missouri
one thousand dollars per semester twice per year for up to ten semesters (Missouri
Department of Higher Education, 2006). It is highly likely that these same students are
being recruited to private institutions or institutions in other states because of their
academic successes.
2-Year Private Sector
The adoption of a merit aid policy had a significant effect on enrollment across
two statesArkansas and Kentucky. In Arkansas, there was a significant negative effect
of the merit aid adoption on enrollment in the 2-year private sector. Kentucky
experienced a positive, immediate impact on enrollment but there was no significant long
term effect. It should be noted that in the case of Kentucky, there was only one 2-year
private institutions data that were included in this study. That is not to say that there is
only one 2-year private institution in the state, but rather this sector did not consistently
182
report its enrollment data. For a list of all of the institutions that were omitted from this
study due to missing data, please refer to Appendix E.
4-Year Private Sector
The adoption of a merit aid policy had a significant negative impact on enrollment
across the 4-year private sector in two states, Arkansas and West Virginia. The
enrollment in the 4-year private institutions in Arkansas, Florida, and West Virginia
dropped over the long run (TTT) as a result of the adoption of the merit aid policies in
those states. For the majority of states, there was no significant impact on enrollment in
the 4-year private as a result of the scholarship program adoption. There were no states
where there was a significant immediate drop in enrollment as a result of a merit aid
policy adoption. One reason that there was not a significant effect may be that in many
4-year private institutions, there are sufficient institutional grant dollars to recruit high
achieving students. Often, students enroll at 4-year private institutions for their prestige,
regardless of cost.
This study found that there is a significant decreasing trend in enrollment over
time (T) at the 4-year private institutions in Arkansas, Florida, Georgia, Mississippi,
Tennessee, and West Virginia, all other variables constant.
183
Table 74
Impact of Policy Adoption by Sector
Sector
Impact 2-Year Public 4-Year Public 2-Year 4-Year
Private Private
Short Term (TT) AK + KY +
WA +
Long Term A K - M O - A R - A R -
(TTT) L A - FL + F L -
WA + W V -
wv-
+ there was a significant positive impact, there was a significant negative impact.
Another variable that was significant in Arkansas, Florida, Georgia, Kentucky,
Louisiana, Michigan, Mississippi, Nevada, South Carolina, and Tennessee was the
average in state tuition (INSTTUIT), which had a significantly positive effect on
enrollment in the 4-year private institutions. One reason that this may be the case is that
in several states, enrollment in private institutions is allowed as part of the merit aid
grant. Another possible reason that enrollment in private 4-year institutions rises as
tuition increases may be that students begin to view private institutions as a viable
alternative past a certain tuition cost.
Research Question 4 (RQ4): What is the effect of the following three types of
merit aid programs across similar programs by sector and level: (a) Full tuition; (b)
184
Partial tuition; (c) One time payment?
Full Tuition Policies
The analysis of the impact of the policy adoptions in the four sectors across states
with similar merit aid programs shows that the full tuition programs have a significant
impact on the 2-year private and 4-year pubic sectors. There was no significant model
for the 2-year public or the 4-year private sectors.
The 2-year private sector enrollment significantly decreases over the long term
(TTT) as a result of the full tuition policy adoption across all full tuition states in this
study. One reason for this may be that the merit aid programs educate students about
their alternatives before they select an institution. Another reason may be that these
institutions are not eligible to receive merit aid money so students that receive the grants
select institutions in either the public or the 4-year private sectors.
There was a significant short term (TT) increase in enrollment in the 4-year public
sector as a result of the adoption of a full tuition payment across the states in this study.
One reason for this may be that the publicity of the programs increases awareness, which
encourages students to go to college. However, the momentum is not sustained, as there
was not a significant long term effect (TTT).
Partial Tuition Policies
The states with partial tuition programs had a significant positive long and short
term enrollment effect at the 2-year public sector across the states. In other words, the
effect of the policy adoption in the partial tuition states was immediate and long term in
the 2-year public sector. It is likely that the 2-year public sector is where students, who
otherwise would not have considered college before the merit aid policy adoption,
185
matriculated. The publicity that surrounds the adoption of merit aid policies in states can
encourage first generation students and parents to think about and plan for college. A
marginal payment may not encourage people to believe they can afford a 4-year
institution, but it is enough for them to be encouraged to attend a 2 year. That there were
significant immediate and long terms effects on enrollment in the 2-year sector after
merit aid adoption supports this suggestion.
One Time Payment Policies
In Michigan, the one state with a one time payment program, there were no
significant effects of the adoption of the merit aid policy in any of the sectors across the
state. It seems as though this supports the assertion that the combination of amount of
award ($2500) and length of time of support (one year) do not serve as an effective
incentive for participation in postsecondary education. Another possible explanation is
that students are enrolling at the next most prestigious level of institution and new
students are entering the pipeline through 2-year public institutions so that the end result
is no significant changes in enrollment in any of the sectors across the state. Table 78
(below) summarizes the significant findings concerning the impact of the three merit aid
policy types across states by sector.
186
Table 75
Impact of Policy Type by Sector
Sector
Policy Type 2-Year Public 4-Year
Public
2-Year 4-Year
Private Private
Full Tuition Program Short Term+ Long Term-
Partial Tuition Program
One Time Payment
Tuition Program
Long Term+
Short Term+
Short Term-
+ denotes a significant positive effect, - denotes a significant negative effect.
Comparison with Existing Literature
Results of this study were largely consistent with the research literature on the
impact of merit aid on postsecondary student enrollment. As in other previously reported
research, state wide adoption of a merit aid policy had a significant effect on student
enrollment (Heller, 1999; Farrell, 2004). Much of the existing research on merit aid is
devoted to understanding how state policies effect student enrollment by group, systems,
and institutions. This study provides a state level examination across states and across
similar merit aid programs thereby expanding the work done by Farrell (2004), Dynarski
(2003) and Heller (1999), among others. Farrell conducted an evaluation of the merit aid
programs in her 2004 dissertation and recommended further examination at the sector
187
level. Dynarski conducted a state level examination of the merit aid programs in her
2003 study, but limited her study to those in the Southern Regional Education Board
(SREB) states. Hellers 1999 study examined the impact of merit aid studies using a
cross sectional time series analysis of the states that had the policies at that time. This
current study includes more states than any of the previous studies and examines the
impact one level deeper by including a sector level analysis.
Previous analyses of the impact of merit aid on sector enrollment found that
students responded to the combination of financial aid along with the tuition increases
that often accompany these policies (Heller, 1999; Pema & Titus, 2004; Kane, 1999).
This study supports those findings, but adds more detail to the general claims of the
previous work, as demonstrated by the long and short term increase in enrollment at the
2-year public institutions with the adoption of partial merit aid programs.
McLendon, Heller, and Young (2006) as well as Cohen-Vogel and Ingle (2006)
conducted examinations of how these state policies transfer across state lines and how
they come to be adopted in neighboring states. However, merit aid policy transfer is not
limited to crossing just state borders; Canada now has a Millennium Scholarship similar
to the merit aid programs in the United States. Canada is not alone; Heller and Rogers
(2004) present implications of the policy transfer of similar higher education policies in
the European Union. Research on how these policies impact regions, institutions, and
students will become increasingly more important as merit aid grows in its utility as a
policy tool across the world. While overall, merit aid policies do increase enrollment
across states, the impact of the merit aid on institutions and sectors can vary depending
on the type of policy that is adopted.
188
Implications for Policy
The context of the state may override merit aid policy adoptions as in Arkansas
and Nevada. However, there is evidence that long-term effects on enrollment are
positive, especially at the broad state level. Questions remain concerning increased
enrollment for whom. The impact of merit scholarships on minority enrollment has been
examined by several researchers with the preponderance of evidence finding that
underrepresented students in higher education remain underrepresented in the merit
scholarship programs (Farrell, 2004; Cornwell & Mustard, 2002, 2004; Dynarski, 2002;
Binder, Ganderton, & Hutchens, 2002). Also, if policy makers want to utilize a certain
sector for cost savings, there is evidence that this may have some credibility. For
example, if states wish to encourage citizens to go to community colleges because of
fiscal austerity, they may be able to deemphasize expensive university enrollment. This
study shows that a partial payment plan encourages two-year enrollment and can
therefore offer a viable alternative to legislators. It is important to note, however, that a
full payment plan may not have the same effect on shifting enrollment in the same
manner, across sectors.
This study also provides a sector level examination across and within states such
that policy makers and higher education administrators would be able to use the models
to predict the changes in enrollment in their state or sector. It is important that state
legislators understand the impact of these policies at the sector and state level for two
main reasons. The first is that as legislators determine whether or not to adopt or adapt
their merit aid policies, they need to consider the higher education capacity in their state.
If policy makers adopt policies that exclude private sectors, this study shows that there
189
can be a negative impact on the private sector institutions in the state. Also, there is
reason to believe that despite the type of merit aid policy, the amount of the award
matters, as demonstrated by Washington and Michigan. In other words, not only should
the policy provide incentive for the first year, but it appears as though the amount of
support through the later years also matters.
Higher education administrators need to be aware of how these policies impact
their institutions and be able to prepare for changes that may arise as a result of the
adoption of a merit aid policy; this study helps them to do that. By providing models
that administrators can utilize in their states to predict enrollment changes, administrators
can allocate resources to accommodate increases in enrollment. Additionally, i f the
model predicts that there will be a decrease in a sector as a result of the adoption;
administrators can play an active role in determining the sectors that are eligible in the
allocation of the awards.
Implications for Future Research
The states in which there was no significant impact of merit aid on enrollment
bring to light some of the shortcomings of this pooled and time series regression. First,
by using self reported secondary data, several issues arise. The data are at the discretion
of the person or institution that is doing the reporting. Since these data cannot be verified
by an outside source, there is no way to determine if the numbers are accurate for any
given year. Additionally, because reporting of these data was not required until 1992,
there were several institutions that did not report their early data (NCES, 2006). Finally,
institutions that did not have Program Participation Agreements (PPAs) with the U.S.
Department of Education did not have to complete the survey in any year. These
190
agreements are in place for institutions that receive federal funding under Title IV of the
Higher Education Act. As a result of these data issues, this study may not fully capture
the impact of the merit aid adoptions on private institutions since the institutions in this
sector made up the bulk of the un- or underreported data.
In the case of Georgia, there was a single year of data (2000) that was an outlier
that was not removed from the analysis. Since these data were not smoothed, this outlier
had a larger effect on the regression line than it would if these data year were discarded
or smoothed. A follow up analysis of the smoothed state data using the centered moving
average revealed that the impact of time (T) and unemployment (UNEMP) on enrollment
(ENROLL) were both significantly positive. This analysis shows that by maintaining the
outlier in the analysis, even smoothed, there is a still a significant negative impact of the
adoption on enrollment in Georgia.
Another issue with this study is that because there were high correlations between
the independent variables, there may be multicollinearity. For example, there is a
consistently high correlation between time (T) and the long term effect variable (TTT)
due to the way that they were coded. Because of the nature of this study, measuring
change over time, it was necessary to include both variables despite the multicollinearity
that may exist.
There are several ways in which this study can be improved upon and expanded.
The first is that by having only three broad characterizations of the merit aid policy, the
estimate of the effects on enrollment are overly generalized. The exact amount of the
award as a predictor of enrollment might provide a clearer picture of the enrollment
response. A case can also be made for including the criteria of the award programs as an
191
indicator of enrollment, similar to the study done by Heller (2004). A study that
examines the programs by selection criteria or award amounts would be useful for
determining how these factors influence enrollment in merit aid states. Another useful
examination might be to characterize the merit programs by amounts rather than payment
types, which would lead to a finer differentiation of programs and their effects.
One important aspect of research on merit aid is what happens to students once
they are enrolled. This study does not include students beyond their first year, but a more
thorough examination of student retention, transfer, and remedial course taking would
provide even more insight as to the student response to state policies. St. John and
Starkey (1995) and St. John (2004) conducted earlier studies of the persistence of
students receiving financial aid but a study of the year to year retention throughout the
merit aid states has yet to be thoroughly examined.
In order to better understand the choices that students are making in response to
merit aid policies, one of the best places to turn for information is the students
themselves. An additional recommendation to improve this research is to conduct a
qualitative study of how and when students generally find out about the programs, how
they think about selecting a college once they decide they want to try to get a merit
scholarship, how they perceive their institutional and sector level choices, and if there is a
strategy about how they will take advantage of the opportunity. A similar study could be
done of university administrators from the various sectors to determine i f there are
differences in how they perceive merit aid programs and the impact on their institutions
and sectors.
192
REFERENCES
Ackerman, R.L., Young, M. & Young, R. (2005). A state-supported, merit-based
scholarship program that works. Journal o f Student Financial Aid, 35 (3).
Acosta (2001). How do colleges respond to changes in federal student aid? Working
Paper #808, Department of Economics, University of California, Los Angeles.
Arkansas Department of Higher Education (2006). Academic Challenge Scholarship
rules and regulations, downloaded from
http://www.arkansaschallenge.com/pdfs/ACS Rules June 2005.pdf
Binder, Ganderton, & Hutchens (2001). Who benefits from a lottery-funded college
subsidy? Evidence from the New Mexico Success Scholarship. Social Science
Research Network.
Bingham & Felbinger (2002). Evaluation in practice: A methodological approach. New
York: NY, Chatham House Publishers of Seven Bridges Press, LLC.
Buss, C., Parker, J. & Rivenburg, J. (2004). Cost, quality and enrollment demand at
liberal arts colleges. Economics of Education Review, 23, 57-65.
Cohen-Vogel, L. & Ingle, K. (2006). When a states neighbors matter most: Diffusion
and postsecondary policy adoption. Paper presented at the annual meeting of the
American Educational Research Association, San Francisco, CA.
Cornwell, C. and Mustard, D. (2003). Race and the effects of Georgias HOPE
Scholarship. In D.E. Heller & P. Marin (Eds.), Who should we help? The
negative consequences o f merit scholarships (pp. 57-72). Cambridge, MA: The
243
Civil Rights Project, Harvard University. Retrieved from
www.law.harvard.edti/civilrights
Cornwell, C., Mustard, D., & Sridhar, D. (2003). The enrollment effects o f merit-based
financial aid: Evidence from Georgia's HOPE Scholarship. Athens, GA:
University of Georgia, Terry College of Business, Economics Department.
Doyle, W. R., Delaney, J.A., & Naughton, B. A. (2004). Institutions amplifying state
policy, Change (July/August), 36-41.
Dynarski, S. (2001). Does aid matter? Measuring the effects of student aid on college
attendance and completion. John F. Kennedy School of Government Faculty
Research working paper, Harvard University.
Dynarski, S. (2002). Race, income, and the impact of merit aid. In D.E. Heller & P.
Marin (Eds.), Who should we help? The negative consequences o f merit
scholarships (pp. 73-91). Cambridge, MA: The Civil Rights Project, Harvard
University. Retrieved from www.law.harvard.edu/civilrights
Dynarski, S. (2003). The new merit aid. Boston, MA: Harvard University, Kennedy
School of Government and NBER.
Education Commission of the States, (2005). Merit scholarships. Retrieved from
http://www.ecs.org/clearinghouse/61/40/6140.htm on November 25, 2005.
The Eleanor Roosevelt papers: The human rights years, 1945-1962. An online historical
site located at George Washington University. Retrieved on August 7, 2005 from
http://www.gwu.edu/~erpapers/abouteleanor/q-and-a/glossarv/nya.htm
Epple, D., Romano, R., & Sieg, H. (2002). On the demographic composition of colleges
and universities in market equilibrium. The American Economic Review, 92(2)
244
[Papers and proceedings of the one hundred fourteenth annual meeting of the
American Economic Association, Atlanta, GA, January 4-6, 2002 (May 2002),
310-314.
Farrell, P.L. (2005). An evaluation of the effectiveness of state non-needs merit-based
scholarship programs (Doctoral dissertation, Michigan State University, 2005).
Dissertation Abstracts International, 65(09), 3299.
FinAid for Educators and FAAs. History of financial aid page. Retrieved on August 7,
2005 from http://www.finaid.org/educators/history.phtml
Gladieux, L. E. (2003). Student assistance the American way. Washington, DC:
Educational Policy Institute, Inc. Retrieved on August 7, 2005 from
http://www.educationalpolicy.org/pdf/Student Assistance.pdf
Gladieux, L. E. & Pema, L. W. (2005). Borrowers who drop out: A neglected aspect o f
the college student loan trend. The National Center for Public Policy and Higher
Education, Report #05-2.
Gladieux, L. E. & Swail, W. S. (1999). Financial aid is not enough. In: King, J.E. (ed.),
Financing College Education: How it works, How i t s Changing. American
Council on Education. Oryx Press: Westport, CT.
Heller, D. E. (2004). The devil is in the details: An analysis o f eligibility criteria for
merit scholarships in Massachusetts. In D.E. Heller & P. Marin (Eds.), State
merit scholarship programs and racial inequality. Cambridge, MA: The Civil
Rights Project at Harvard University.
Heller, D.E. (Jul/Aug2004). The changing nature o f financial aid. Academe, 90(4), 36-
39.
245
Heller, D. (2003). Merit scholarships and incentives f o r academic performance. Paper
presented at the annual conference of the Association for the Study of Higher
Education in Portland, OR, November 2003.
Heller, D. (2002) State merit scholarship programs: An introduction. In D.E. Heller & P.
Marin (Eds.), Who should we help? The negative consequences o f merit
scholarships (pp. 17-24). Cambridge, MA: The Civil Rights Project, Harvard
University.
Heller, D.E. (1999). The effects of tuition and state financial aid on public college
enrollment. Review o f Higher Education, 23,65-89.
Heller, D.E. (1997). Student price response in higher education: An update to Leslie and
Brinkman. Journal o f Higher Education, 68, 624-659.
Heller, D.E., & Marin, P. (Eds.) (2004). State merit scholarship programs and racial
inequality. Cambridge, MA: The Harvard Civil Rights Project at Harvard
University.
Heller, D.E., & Marin, P. (Eds.). (2002). Who should we help? The negative social
consequences o f merit scholarships. Cambridge, MA: The Civil Rights Project at
Harvard University.
Heller, D.E. & Rasmussen (2002). Merit scholarships and college access: Evidence from
Florida and Michigan. In D.E. Heller & P. Marin (Eds.), Who should we help?
The negative consequences o f merit scholarships (pp. 17-24). Cambridge, MA:
The Civil Rights Project, Harvard University.
Heller, D.E. & Rogers, K. (2004). Shifting the burden: Public and private financing o f
higher education in the United States and implications for Europe. Paper
246
presented at the European Higher Education Society 26th Annual Forum, in
Barcelona, Spain, September 2004.
Heller, D.E. & Rogers, K. (2003). Merit scholarships and incentives f o r academic
performance. Paper presented at the Association for the Study of Higher
Education, Portland, Oregon, November, 2003.
Herzog, S. (2005). Measuring the impact of state-funded merit scholarship programs on
student access and success: Evidence from the nations fastest growing state.
Presentation at the Association for Institutional Research Forum, San Diego, CA,
May 29-June1, 2005.
Holmes, M. D., Daudistel, H. C. & Taggart, W. A. (1992). Plea bargaining policy and
state district court caseloads: An interrupted time series analysis. Law & Society
Review, 26(1), 139-159.
Institute for Higher Education Policy (2005). Private scholarships count: Access to
higher education and the critical role of the private sector. Paper presentation at
the 22nd Annual Student Financial Aid Research Network Conference, June 9 -
11, 2005, Chicago, IL
Institute for Higher Education Policy (2005). How Latino students pay for college:
Patterns in Financial Aid, 2003-04. Retrieved from
http://www.ihep.org/Pubs/PDF/LatinoEng.pdf
Jackson, G. A. (1978). Financial aid and student enrollment. The Journal o f Higher
Education, 49(6), 548-574.
Kane, T.J. (1999). The price o f admission: Rethinking how Americans pay f o r college.
Washington, D.C., Brookings Institution Press.
247
Kane, T. J. (2003). A quasi-experimental estimate of the impact of financial aid on
college-going. National Bureau of Economic Research Working Paper 9703.
Kim, D. (2004). The effect of financial aid on students college choice: Differences by
racial group. Research in Higher Education, 45(1).
Klein, J. (1997). Statistical visions in time: A history o f time series analysis, 1662-1938.
New York, NY: Cambridge University Press.
Lewis-Beck, M. (1980). Applied regression: An introduction. Series: Quantitative
applications in the social sciences, No. 22. Newbury Park, CA: SAGE
Publications.
Long, B. T. (2003). How do financial aid policies affect colleges? The institutional
impact of the Georgia HOPE Scholarship. Harvard Graduate School of
Education and NBER (July 15, 2003 version).
Maag & Fitzpatrick (2004). Federal financial aid for higher education: Programs and
prospects. Report for the Urban Institute.
Martinez, M. (2004).Postsecondary participation and state policy. Sterling, VA: Stylus
Publishing.
McDowall, D., McLeary, R., Meidinger, E.E., & Hay, Jr. R.A. (1980). Interrupted time
series analysis. Series: Quantitative applications in the social sciences, No. 21.
Newbury Park, CA: SAGE Publications
McLendon, M.K., Heller, D.E., & Young, S.P. (2005). State postsecondary policy
innovation: Politics, competition, and the interstate migration of policy ideas. The
Journal o f Higher Education, 76 (4).
248
Meier, K.J. & Brudney, J.L. (2002). Applied statistics for public administration.
Orlando, FL. Harcourt, Inc.
Meyer, B. D. (1995). Natural and quasi-experiments in economics. Journal o f Business
and Economic Statistics, 13(2).
Missouri Department of Higher Education (2006). Missouri Higher Education Academic
"Bright Flight" Scholarship Program information sheet retrieved from
http://www.dhe.mo. gov/hsstudentsbri ghtfli ght.shtml on July 17, 2006.
Mortenson, T. (2001). College participation by gender age 18 to 24, 1967 to 2000,
Postsecondary Education OPPORTUNITY Newsletter. Number 109, July 2001.
National Center for Education Statistics (2006). Introduction to the Integrated
Postsecondary Division. Retrieved from
http://165.224.221.98/ipeds/AboutIPEDS.asp on July 12, 2006.
National Center for Higher Education Management Systems (NCHEMS) (2005).
Retrieved from:
http://www.hi gheredinfo.org/dbrowser/index.php?submeasure=63&year=2002&l
evel=nation&mode=data&statc=0 on November 27, 2005.
Ostrom, C. W. Jr. (1990). Time series analysis: regression techniques. Series:
Quantitative applications in the social sciences, No. 9. Thousand Oaks, CA:
SAGE Publications.
Pema, L. W. & Titus, M. (2004). Understanding differences in the choice of college
attended: The role of state public policies. The Review o f Higher Education,
27(4), pp. 501-525.
249
Ruppert, S. (2003). Closing the college participation gap: A national summary.
Education Commission of the States Report.
Sayrs, L. (1989). Pooled time series analysis. Series: Quantitative applications in the
social sciences, No. 70. Newbury Park, CA: SAGE Publications.
Smart, J. (2005). Attributes of exemplary research manuscript employing quantitative
analyses. Research in Higher Education, 46 (4), pp. 461-477.
South Carolina Commission on Higher Education (2006) Details outlining the
scholarship guidelines, Retrieved on July 14, 2006 from
http://www.che.sc.gov/New_Web/ForInstitutions/Scholarship_Guidelns.htm
StatSoft (2005) Electronic statistics book. Retrieved from
http://www.statsoft.com/textbook/sttimser.html, on November 25, 2005.
Stewart, M. A. & Post, P. (1990). Minority students perceptions of variables affecting
their selection of a large university. Journal o f Multicultural Counseling &
Development, 18(4).
St. John, E. P. (1994). Prices, productivity, and investment: Assessing financial
strategies in higher education. ASHE-ERIC Higher Education Report No. 3.
Washington, DC: The George Washington University.
St. John, E. P. (1991). The impact of student financial aid: A review of recent research.
Journal o f Student Financial Aid, 21(1), 18-32.
St. John, E. P. & Starkey, J. B. (1995) An alternative to net price: Assessing the influence
of prices and subsidies on within-year persistence. The Journal o f Higher
Education, 66(2), 156-186.
250
Stream, C. (1999). Health reform in the states: A model of state small group health
insurance market reforms. Political Research Quarterly, 52(3), 499-525.
Taggart, W. (1989). Redefining the power of the federal judiciary: The impact of court-
ordered prison reform on state expenditures for corrections. Law & Society
Review, 23(2).
Thistlethwaite, D. L. (1958). College-scholarship offers and the enrollment of talented
students. The Journal o f Higher Education, 29(8), 421-424, 467-468
Thomas, S. L. (2003). Longer-term economic effects of college selectivity and control.
Research in Higher Education, 44(3).
Tierney, M. L. (1980). The impact of financial aid on student demand for public/private
higher education. The Journal o f Higher Education, 51 (5), 527-545
Van der Klaauw, W. (2002). Estimating the effect of financial aid offers on college
enrollment: A regression-discontinuity approach. International Economic
Review, 43(4), 1249-1287.
Washington Higher Education Coordinating Board (2006). Retrieved from
http ://www. hecb. wa. gov/financial aid/ wps/wpsindex. asp
251
VITA
Graduate College
Local Address:
P.O. Box 70781
Las Vegas, Nevada 89170
Home Address:
9701 Fish Lake Road
Holly, Michigan 48442
Degrees:
Bachelor of Arts, Biological Sciences, 1998
Wayne State University, Detroit
Master of Arts, Educational Management and Development, 2003
New Mexico State University, Las Cruces
Publications:
Martinez, M.C. & Nilson, M.J. (2006). Assessing the connection between higher
education policy and performance, Educational Policy.
Dissertation Title: Evaluating the Effect of Merit Aid as a Higher Education Policy
Tool Using Time Series Analysis
Dissertation Examination Committee:
Chairperson, Dr. Mario C. Martinez
Committee Member, Dr. Robert Ackerman
Committee Member: Dr. Mimi Wolverton
Graduate Faculty Representative: Dr. Christopher Stream, Ph.D.
252

Note To Users

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Note To Users

Uploaded by

Copyright:

Available Formats

NOTE TO USERS

Page(s) not included in the original manuscript and are

You might also like